AI and Model Improvement Notice

This notice explains how VectorPitch uses artificial intelligence and automated analysis, how AI Outputs should be treated, and how — and with what safeguards — we use footage and data to improve our models. For AI or data protection enquiries, email privacy@vectorpitch.com.

Last updated: 2026-10-18 · v1.0

Version: 1.0
Effective Date: 2026-10-18
Company: Vector Sport Solutions Limited
Privacy Contact: privacy@vectorpitch.com
Security Contact: security@vectorpitch.com
Model Improvement Plan Information: See Schedule E of this Notice.

Clause

Provision

Section 1 — Status and Scope of this Notice

On AI matters this Notice comes first where it clashes with the main Terms, but the Data Processing Agreement, Privacy Notice and data protection law still win where they set a stricter rule. Reading this Notice doesn't switch model improvement on — if your plan or the law excludes it, it stays excluded.

1.1

This AI and Model Improvement Notice explains how VectorPitch uses artificial intelligence, automated analysis and model improvement in connection with the Service.

1.2

This Notice should be read together with the Terms of Service Agreement, Data Processing Agreement, Privacy Notice, Acceptable Use and Customer Content Policy, Sub-processor List and any applicable Subscription Plan.

1.3

This Notice applies to AI Features made available through the Service and to model evaluation and improvement carried out by VectorPitch.

1.4

Where this Notice specifically addresses AI Features, AI Outputs, Model Improvement or AI-related data processing, this Notice will prevail over the Terms of Service Agreement to the extent of any inconsistency on those subjects. The Terms of Service Agreement will continue to govern contractual matters not specifically addressed in this Notice. Where the Data Processing Agreement, Privacy Notice or Applicable Data Protection Laws impose stricter requirements than this Notice, those requirements will prevail.

1.5

This Notice is intended to provide transparency about AI Features, AI Outputs, human review, Model Improvement, data use, safeguards, third-party AI providers, customer responsibilities and plan-level Model Improvement controls.

1.6

This Notice does not itself authorise VectorPitch to process Customer Content for Model Improvement where such processing is excluded by the applicable Subscription Plan, product control, written agreement or Applicable Data Protection Laws.

1.7

Any plain-English explanations, summaries, panels, side notes or similar materials displayed alongside this Notice are provided for convenience only. They do not amend, limit or override the legal terms set out in this Notice.

Section 2 — Definitions

Terms used throughout. The key one is De-identified Training Material: before footage is used for training we mask faces, remove audio and strip names, teams and match dates. We train on that processed material, not the original video.

2.1

In this Notice:

AI Features means features that use automated analysis, machine learning, computer vision, artificial intelligence or related technologies to generate tags, tracking, analysis, confidence scores, events, reports or other outputs.

AI Outputs means outputs generated or assisted by AI Features, including tags, tracking data, event detections, confidence scores, annotations, reports, tactical outputs, player movement metrics, summaries, recommendations or other analysis.

Applicable AI Laws has the meaning given in the Terms.

Applicable Data Protection Laws has the meaning given in the Privacy Notice.

Authorised User has the meaning given in the Terms.

Customer Content has the meaning given in the Terms.

Customer Personal Data has the meaning given in the Data Processing Agreement.

De-identified Training Material means frames, crops, derived tracking data, tags, corrections, annotations or related outputs prepared for model evaluation or improvement in a manner designed to remove direct identity linkage, including by masking faces, removing audio and removing, severing or separating match, date, organisation, workspace, roster and direct player identity metadata using appropriate technical and organisational measures.

High-Risk Use has the meaning given in the Terms.

Model Improvement means evaluating, testing, training, fine-tuning, validating, improving or developing AI Features, models, systems, datasets, workflows or performance metrics used by VectorPitch.

Service means VectorPitch and all related products, features, applications, websites, APIs, software, hosted services and support services made available by VectorPitch.

Special Category Data has the meaning given in the Privacy Notice.

VectorPitch, we, us or our means Vector Sport Solutions Limited.

Section 3 — AI Processing Roles

Creating analysis inside your workspace, we handle data for you and on your instructions, under the Data Processing Agreement. Using data to improve our own models, we act for our own purposes under this Notice and the Privacy Notice instead.

3.1

VectorPitch processes AI-related data in different legal roles depending on the processing context.

3.2

Where VectorPitch generates AI Outputs for a customer within the customer's workspace, VectorPitch normally acts as processor and processes Customer Personal Data on the customer's documented instructions, as described in the Data Processing Agreement.

3.3

Where VectorPitch processes material for Model Improvement for VectorPitch's own purposes, VectorPitch acts as independent controller, as described in the Privacy Notice and this Notice.

3.4

VectorPitch's role as independent controller for Model Improvement does not make VectorPitch the controller of all Customer Content in the customer's workspace.

3.5

Customers remain responsible for their own use of Customer Content, AI Features and AI Outputs, including their own lawful basis, transparency notices, safeguarding obligations, AI law obligations and decisions made using or informed by AI Outputs.

Section 4 — Intended Purpose of AI Features

The AI is built for sports analysis — not for medical, safeguarding, employment, biometric or automated-decision use, and it assists people rather than replacing them. Not every sport gets every feature; some may have few or none, depending on what we've built and your plan.

4.1

VectorPitch provides AI Features for sports video analysis.

4.2

The intended purpose of AI Features is to support coaching, tactical analysis, video review, performance analysis and sports decision-support.

4.3

Depending on the applicable Subscription Plan and available functionality, AI Features may assist with:

(a) detecting players;
(b) detecting the ball;
(c) detecting pitch features;
(d) analysing camera type, camera angle and video format;
(e) generating player tracking;
(f) generating movement, distance, speed, acceleration, positioning and related sports metrics;
(g) generating automated tags;
(h) detecting events;
(i) identifying patterns in play;
(j) generating heatmaps, passing networks, tactical outputs or similar analysis;
(k) generating confidence scores;
(l) generating reports, summaries or analysis outputs; and
(m) supporting clips, playlists, annotations, presentations and exports.

4.4

AI Features and the capabilities described in clause 4.3 are not available for every sport. The AI Features available for a particular sport depend on the applicable Subscription Plan, product configuration and the AI Features that VectorPitch has developed, validated and enabled for that sport. For some sports only a limited subset of AI Features may be available, and for some sports no AI Features may be available.

4.5

AI Features are not intended for medical, healthcare, injury, safeguarding, employment, disciplinary, contractual, scholarship, eligibility, biometric identification, biometric categorisation, emotion recognition or automated decision-making purposes.

4.6

AI Features are not intended to replace professional judgement by coaches, analysts, club staff, academy staff, medical professionals, safeguarding personnel, educators, employers or other responsible humans.

Section 5 — AI Outputs Are Decision-Support Only

Treat AI outputs as input to a human decision, not the decision itself, and check and correct them before you act.

5.1

AI Outputs are provided for decision-support only.

5.2

AI Outputs may be incomplete, inaccurate, delayed, inconsistent, unsuitable for a particular purpose or affected by the quality and nature of input footage.

5.3

Analysis quality may vary depending on factors including:

(a) camera angle;
(b) video resolution;
(c) frame rate;
(d) lighting;
(e) weather;
(f) pitch visibility;
(g) player visibility;
(h) occlusion;
(i) camera movement;
(j) kit similarity;
(k) shadows;
(l) crowding of players;
(m) broadcast overlays;
(n) video compression;
(o) whether the footage is broadcast, panoramic, fixed camera, highlights or training footage; and
(p) whether the footage includes unusual pitch markings, non-standard formats or partial pitch coverage.

5.4

Customers and Authorised Users are responsible for reviewing, correcting and validating AI Outputs before relying on them.

5.5

AI Outputs must not be used as the sole basis for any High-Risk Use.

5.6

The Service may display confidence scores, quality indicators or other information designed to help users understand the likely reliability of certain AI Outputs. Such information is indicative only and is not a guarantee of accuracy.

5.7

VectorPitch does not warrant or guarantee the accuracy, completeness, reliability, suitability or availability of AI Outputs except to the extent expressly stated in a separate written agreement.

Section 6 — Human Review and Customer Responsibility

For any decision that materially affects a person, the AI can inform it — but a qualified human must make and own it.

6.1

Customers must maintain appropriate human review and oversight when using AI Features and AI Outputs.

6.2

Customers are responsible for determining how AI Outputs are interpreted, validated, corrected, shared and used within their organisation.

6.3

Customers must ensure that Authorised Users are appropriately trained, supervised and authorised to use AI Features and AI Outputs.

6.4

Customers must not present AI Outputs as verified human analysis unless they have been reviewed and validated by a suitably qualified human.

6.5

Customers must not use AI Outputs as the sole basis for decisions relating to:

(a) player health;
(b) injury;
(c) medical treatment;
(d) safeguarding;
(e) employment;
(f) discipline;
(g) player selection;
(h) contracts;
(i) scholarships;
(j) eligibility;
(k) player welfare;
(l) access to opportunities; and
(m) any other decision that may produce legal effects or similarly significant effects concerning an individual.

6.6

Customers are responsible for complying with Applicable AI Laws, Applicable Data Protection Laws, safeguarding obligations, league rules, competition rules, employment laws, education laws and other legal or regulatory requirements that apply to their use of the Service and AI Outputs.

Section 7 — No Solely Automated Decisions by VectorPitch

We don't make automated decisions about people.

7.1

VectorPitch does not use AI Features to make decisions about individuals that produce legal effects or similarly significant effects.

7.2

VectorPitch does not decide whether a player is selected, disciplined, contracted, retained, promoted, released, medically assessed, injured, safeguarded, eligible or suitable for an opportunity.

7.3

Customers are responsible for any decisions they make using or informed by AI Outputs.

7.4

If a customer uses AI Outputs in a way that produces legal effects or similarly significant effects concerning an individual, that use is outside VectorPitch's intended purpose unless expressly agreed in writing with VectorPitch and permitted by applicable law.

Section 8 — AI Act and Regulatory Classification

Used as intended, for sports analysis, we assess the AI Features as not high-risk under the EU AI Act. If you use the service for a high-risk purpose, the duties that come with that are yours.

8.1

VectorPitch's intended purpose is sports coaching, tactical analysis, video review and decision-support.

8.2

VectorPitch does not provide the Service as a medical device, employment decision system, recruitment system, worker management system, education admissions system, disciplinary decision system, biometric identification system, biometric categorisation system, emotion recognition system or automated legal decision system.

8.3

VectorPitch has assessed, or will assess before launch, the AI Act classification of its launch AI Features based on their intended purpose, design, functionality, marketing, instructions for use, contractual restrictions and applicable usage controls.

8.4

VectorPitch's current position is that the launch AI Features are not high-risk AI systems under Annex III of the EU AI Act when used for their intended purpose of sports coaching, tactical analysis, video review and decision-support.

8.5

The position in clause 8.4 is based on the following factors:

(a) the Service is not designed, marketed or instructed for employment decision-making, recruitment, education admissions, biometric identification, biometric categorisation, emotion recognition, medical assessment, law enforcement or other High-Risk Use;
(b) VectorPitch does not use AI Features to make decisions that produce legal effects or similarly significant effects concerning individuals;
(c) AI Outputs are provided for decision-support and require appropriate human review;
(d) the Terms prohibit use of AI Outputs as the sole basis for High-Risk Use;
(e) the Service is intended to support sports coaching, tactical analysis and video review rather than to determine employment, contractual, disciplinary, medical, safeguarding, scholarship, eligibility or welfare outcomes.

8.6

VectorPitch will document and keep its AI Act classification position under review as product features, regulatory guidance, applicable AI laws, customer usage patterns and reasonably foreseeable misuse risks develop.

8.7

If VectorPitch determines that an AI Feature is, or becomes, subject to high-risk AI system requirements under Applicable AI Laws for its intended purpose, VectorPitch will take the steps required by those laws before making that AI Feature available for the relevant purpose. Such steps will include, as required by Applicable AI Laws, updates to risk management, documentation, transparency information, human oversight measures, technical measures, contractual terms and other compliance steps mandated for the relevant AI system classification.

8.8

If a customer uses the Service or AI Outputs for a purpose that is high-risk, prohibited, regulated or legally significant under Applicable AI Laws, the customer is responsible for its own compliance obligations as deployer, user, employer, education provider, club, academy, school, controller or other regulated person, as applicable.

8.9

Customers must not use the Service for biometric identification, biometric categorisation, emotion recognition, medical assessment, injury assessment, employment decision-making, recruitment, disciplinary decisions, safeguarding decisions or other High-Risk Use unless expressly agreed in writing with VectorPitch and permitted by applicable law.

8.10

VectorPitch may provide documentation, warnings, usage guidance, confidence indicators, human oversight guidance and other information to help customers use AI Features responsibly. Such information does not transfer the customer's compliance obligations to VectorPitch.

8.11

If a customer materially changes the intended purpose of the Service or integrates AI Outputs into a regulated decision-making workflow, the customer is responsible for assessing whether that change creates additional legal obligations under Applicable AI Laws.

Section 9 — AI Transparency

We label AI-generated and AI-assisted outputs in the service where required, and where practical show confidence scores and quality indicators alongside them.

9.1

VectorPitch will label or otherwise identify AI-generated or AI-assisted outputs in the Service where required by Applicable AI Laws and, more generally, where reasonably necessary to support transparency, human review and responsible use.

9.2

Where applicable law requires disclosure that an output has been AI-generated or AI-assisted, VectorPitch will implement appropriate labelling or notification measures.

9.3

Where practical and appropriate, AI Outputs may include:

(a) confidence scores;
(b) quality indicators;
(c) labels identifying AI-generated or AI-assisted content;
(d) warnings about experimental or beta functionality;
(e) explanations or supporting information for certain output types; and
(f) prompts for human review or correction.

9.4

Not all AI Outputs will include the same level of explanation, confidence scoring or interpretability.

9.5

A confidence score is not a guarantee that an AI Output is correct.

9.6

Customers are responsible for providing any notices required to players, staff, parents, guardians, opposition teams or other individuals in relation to the customer's use of AI Features and AI Outputs.

Section 10 — Beta, Experimental and Preview AI Features

Beta features are less reliable — review their output more carefully.

10.1

AI Features may be released as beta, preview, trial, early-access or experimental features.

10.2

Beta, preview, trial, early-access and experimental AI Features may be incomplete, inaccurate, unreliable, unavailable, modified, restricted or withdrawn.

10.3

Such features are provided for evaluation and feedback unless otherwise stated.

10.4

VectorPitch may change or discontinue beta, preview, trial, early-access or experimental AI Features at any time.

10.5

Processing times are not guaranteed unless expressly stated in a separate written agreement.

10.6

Customers must apply heightened human review when using beta, preview, trial, early-access or experimental AI Features.

Section 11 — Corrections, Feedback and Error Reporting

When you fix or annotate an output, that feedback may help us improve the service.

11.1

Authorised Users may be able to view, edit, correct, annotate, confirm or reject AI Outputs.

11.2

Corrections, edits, annotations, confirmations, rejections, comments and reported errors may be used to provide, maintain, evaluate, improve and develop the Service, including AI Features.

11.3

Customer-reported errors may be used to identify issues, improve product quality, improve documentation, improve model performance and reduce similar errors in the future.

11.4

Corrections, annotations, confirmations, rejections and reported errors are considered in VectorPitch's LIA and DPIA for Model Improvement where they are used for Model Improvement involving personal data.

11.5

Customers and Authorised Users are encouraged to report incorrect AI Outputs through available product or support channels.

11.6

VectorPitch may review reported errors, support requests or disputed outputs where necessary for support, troubleshooting, quality review, security, legal compliance or service improvement.

Section 12 — Model Improvement Overview

The product works out of the box; your content isn't needed to make it function. Whether your content is used for model improvement depends on your plan and the controls you set.

12.1

The base Service runs on pre-trained models.

12.2

Customer-derived Model Improvement is used to improve accuracy on real-world sports footage, including amateur, academy, youth, school, semi-professional and non-professional footage, varied camera angles, lower production quality video, different kits, different lighting conditions and different playing environments.

12.3

Customer-derived Model Improvement improves the Service but is not required for the base Service to function.

12.4

Depending on the applicable Subscription Plan and available controls, VectorPitch may use Customer Content and related outputs to evaluate, train, fine-tune, test, validate and improve AI Features.

12.5

Each Subscription Plan will state whether Model Improvement is included, excluded, optional or subject to customer control. The current Model Improvement position for each Subscription Plan is set out in Schedule E of this Notice and may also be displayed during sign-up, plan selection, purchase, beta onboarding or upload.

12.6

Where a Subscription Plan includes Model Improvement, VectorPitch will take reasonable steps to disclose that position before the first upload of Customer Content under that Subscription Plan and to keep a record of the relevant acceptance or disclosure event.

12.7

Model Improvement may include improvement of:

(a) player detection;
(b) ball detection;
(c) pitch detection;
(d) camera classification;
(e) tracking;
(f) tag generation;
(g) event detection;
(h) tactical analysis;
(i) quality assessment;
(j) output confidence scoring;
(k) error detection;
(l) report generation; and
(m) other sports analysis functionality.

Section 13 — Material Used for Model Improvement

We don't keep raw identifiable footage in the training set. If something can't be effectively de-identified, we don't train on it. And we don't build facial recognition or individual-identification models from your footage.

13.1

Where Customer Content is used for Model Improvement, VectorPitch does not retain raw identifiable video in the training corpus.

13.2

Model Improvement is designed to use De-identified Training Material. Where material cannot be effectively de-identified in accordance with this Notice, VectorPitch will not use that material for Model Improvement.

13.3

De-identified Training Material may include:

(a) face-masked frames or crops;
(b) selected video-derived frames;
(c) selected image crops;
(d) derived tracking data;
(e) object detection data;
(f) pitch detection data;
(g) ball detection data;
(h) player movement data;
(i) tags;
(j) corrections;
(k) annotations;
(l) confidence information;
(m) derived metadata;
(n) error reports; and
(o) related sports analysis outputs.

13.4

Audio is removed from material used for Model Improvement.

13.5

Match, date, organisation, workspace, roster and direct player identity metadata is removed, severed or separated from material used for Model Improvement using appropriate technical and organisational measures.

13.6

VectorPitch does not use Customer Content to train models to identify named individuals.

13.7

VectorPitch does not create biometric templates for the purpose of uniquely identifying individuals.

13.8

VectorPitch does not use Model Improvement datasets to recognise the same named player across unrelated videos.

13.9

VectorPitch does not use Model Improvement datasets to infer health, injury, ethnicity, religion, political opinion, sexual orientation or other Special Category Data.

13.10

VectorPitch does not use Customer Content in public demos, public marketing materials, case studies or public customer stories without the customer's prior agreement.

13.11

VectorPitch may retain model performance information, aggregate metrics, statistical information, model weights or other outputs derived from Model Improvement where such information does not identify an individual.

Section 14 — Youth, Academy and School Footage

If you use footage of under-18s, the consents, notices and safeguards are yours to put in place. To keep youth footage out of model improvement, pick a plan that excludes it, use the control, or email us — before that footage is used.

14.1

Customer Content may include children or young players where customers use the Service for academies, schools, youth teams or similar programmes.

14.2

Customers are responsible for ensuring that use of youth, academy or school footage is lawful, fair, transparent, appropriately permissioned and properly safeguarded.

14.3

Where youth, academy or school footage is used for Model Improvement, VectorPitch applies the safeguards described in this Notice, including face masking, audio removal, metadata severance, restricted access and use of De-identified Training Material rather than retained raw identifiable video in the training corpus.

14.4

Additional youth, academy or school Model Improvement controls may apply depending on the Subscription Plan, product configuration or written agreement.

14.5

Customers that require exclusion of youth, academy or school footage from Model Improvement must select a Subscription Plan that includes that exclusion, enable the applicable product control if available, or notify VectorPitch at privacy@vectorpitch.com or support@vectorpitch.com, identifying the relevant Customer Content to be excluded, before that content is processed by VectorPitch for Model Improvement purposes.

14.6

VectorPitch's legitimate interests assessment and data protection impact assessment for Model Improvement will specifically consider the possibility that source footage may include children, opposition players, spectators and other individuals who do not have a direct account relationship with VectorPitch.

Section 15 — Opposition Players, Spectators and Incidental Individuals

Footage picks up opposition players, referees, staff, spectators and passers-by. Having the rights, notices and lawful bases for them is your responsibility; our safeguards are designed to reduce how identifiable they are in training material.

15.1

Customer Content may include opposition players, referees, staff, spectators, parents, guardians and other incidental individuals.

15.2

Customers are responsible for ensuring that they have the rights, notices, permissions and lawful bases required for their upload and use of such Customer Content.

15.3

VectorPitch's Model Improvement safeguards are designed to reduce the identifiability of individuals in Model Improvement datasets, including individuals who do not have a direct account relationship with VectorPitch.

15.4

VectorPitch does not use Model Improvement to create profiles of opposition players, spectators or incidental individuals.

15.5

VectorPitch does not use Model Improvement to identify or contact opposition players, spectators or incidental individuals.

Section 16 — Third-Party AI Providers

Your content is never sent to providers such as OpenAI or Anthropic for them to train their models.

16.1

VectorPitch may use third-party AI providers to provide inference, enrichment or other Service functionality where enabled.

16.2

Third-party AI providers may include OpenAI, Anthropic or similar providers where listed in the Sub-processor List and enabled for the relevant feature.

16.3

Customer Content is not sent to OpenAI, Anthropic or similar third-party AI providers for model training.

16.4

VectorPitch does not permit third-party AI providers to use Customer Content to train their own models.

16.5

Where third-party AI providers are used to provide inference, enrichment or other Service functionality, VectorPitch uses appropriate technical, organisational and contractual measures to configure those services so that Customer Content is not used to train the provider's models.

16.6

Where an optional AI Feature uses a third-party AI provider, the relevant provider and processing details will be identified in the Sub-processor List or applicable product documentation.

16.7

Customers may choose not to enable optional AI Features that use third-party AI providers unless the relevant feature is included as a mandatory part of the applicable Subscription Plan.

16.8

If a third-party AI provider is required for a material AI Feature, VectorPitch will identify that provider in the Sub-processor List in accordance with the Data Processing Agreement.

Section 17 — Lawful Basis for VectorPitch Model Improvement

We rely on legitimate interests, the basis commonly used for product improvement. Match footage captures people we have no relationship with — opponents, referees, the crowd — so asking each for consent isn't realistic. We protect them through de-identification instead.

17.1

Where GDPR applies and Model Improvement involves personal data processed by VectorPitch as independent controller, VectorPitch's anticipated lawful basis is legitimate interests under Article 6(1)(f) GDPR.

17.2

VectorPitch's legitimate interests include:

(a) improving sports analysis accuracy;
(b) improving performance on real-world sports footage;
(c) improving performance for amateur, academy, school, youth, semi-professional and non-professional environments;
(d) reducing tagging, detection and tracking errors;
(e) improving reliability across varied cameras, kits, lighting conditions, pitch conditions and match environments;
(f) improving the safety, consistency and usefulness of AI Features;
(g) reducing customer correction burden;
(h) improving product quality; and
(i) developing and maintaining a competitive, reliable and affordable sports analysis service.

17.3

VectorPitch considers this processing necessary for those legitimate interests because public or professional broadcast datasets do not adequately represent many customer environments, including amateur, academy, school, youth, semi-professional, non-professional, fixed-camera, panoramic and lower production quality footage.

17.4

VectorPitch has considered consent as a potential lawful basis for Model Improvement. VectorPitch's position is that legitimate interests is the more appropriate primary lawful basis because:

(a) Customer Content may include many categories of individuals, including account users, non-user players, opposition players, referees, staff, spectators, parents, guardians and incidental individuals;
(b) VectorPitch may not have a direct relationship with many individuals appearing in Customer Content;
(c) obtaining, recording and managing valid consent from all individuals appearing in sports footage would be impracticable in many real-world match contexts, particularly where footage includes opposition players, referees, spectators, parents, guardians and incidental individuals;
(d) the practical and logistical obstacles to obtaining reliable informed consent from all such individuals are substantial;
(e) the processing is subject to safeguards designed to reduce identifiability and impact; and
(f) individuals retain the right to object where personal data remains in scope and legitimate interests is relied on.

17.5

VectorPitch will not rely on legitimate interests where Applicable Data Protection Laws require consent or another lawful basis for the relevant processing.

17.6

VectorPitch has conducted or will conduct a legitimate interests assessment before relying on legitimate interests for Model Improvement involving personal data.

17.7

VectorPitch has conducted or will conduct a data protection impact assessment under Article 35 GDPR in respect of Model Improvement processing. VectorPitch treats Model Improvement from Customer Content as requiring DPIA review due to the nature of the processing, including automated analysis of video footage, possible inclusion of children and other vulnerable data subjects, movement and tracking data, use of novel technology and potential scale of processing.

17.8

VectorPitch applies safeguards designed to reduce the impact of Model Improvement on individuals, including:

(a) face masking;
(b) audio removal;
(c) severance of match, date, organisation, workspace, roster and direct player identity metadata using appropriate technical and organisational measures;
(d) use of De-identified Training Material rather than retained raw identifiable video in the training corpus;
(e) no model training to identify named individuals;
(f) no creation of biometric templates for the purpose of uniquely identifying individuals;
(g) no third-party AI provider model training on Customer Content;
(h) restricted access to Model Improvement datasets;
(i) periodic review of de-identification measures;
(j) objection handling;
(k) plan-level exclusion controls where available;
(l) no use of AI Outputs by VectorPitch to make legal or similarly significant decisions about individuals; and
(m) consideration of children, opposition players, spectators and incidental individuals in the LIA and DPIA.

17.9

Where Special Category Data is intentionally processed for Model Improvement, VectorPitch will not rely solely on Article 6(1)(f) GDPR and will ensure that an Article 9 GDPR condition or equivalent legal basis applies. VectorPitch does not intend Model Improvement to process Special Category Data.

Section 18 — Model Improvement Exclusions

Three ways to opt out: a plan that excludes it, the product control, or an email to us — each before the content is used. An opt-out applies going forward and doesn't undo training material already created.

18.1

Some Subscription Plans may include use of Customer Content and related outputs for Model Improvement.

18.2

Some Subscription Plans may include a Model Improvement exclusion.

18.3

Each Subscription Plan will state whether Model Improvement is included, excluded, optional or subject to customer control.

18.4

Where a customer requires exclusion from Model Improvement involving Customer Content, the customer must:

(a) select a Subscription Plan that includes a Model Improvement exclusion;
(b) enable the applicable product control if available; or
(c) notify VectorPitch at privacy@vectorpitch.com or support@vectorpitch.com, identifying the Customer Content to be excluded, before that Customer Content is processed by VectorPitch for Model Improvement purposes.

18.5

A Model Improvement exclusion is not effective unless confirmed in the Service, stated in the applicable Subscription Plan or confirmed by VectorPitch in writing.

18.6

Unless otherwise agreed, a Model Improvement exclusion applies prospectively only.

18.7

Unless required by Applicable Data Protection Laws, a competent supervisory authority, a court or the applicable Subscription Plan, a Model Improvement exclusion does not require:

(a) deletion of De-identified Training Material already created;
(b) deletion of anonymised material;
(c) retraining of models already developed; or
(d) deletion of statistical or aggregate model performance information.

18.8

If a customer requires Model Improvement exclusion for youth, academy or school content, the customer should confirm that exclusion before that content is processed by VectorPitch for Model Improvement purposes.

Section 19 — Data Subject Objections

Where footage was uploaded by a club we may need to involve that club to act on an objection. Once material is genuinely anonymous it's no longer personal data, so rights like deletion may no longer attach to it.

19.1

Where VectorPitch relies on legitimate interests for Model Improvement involving personal data, data subjects may have the right to object to processing of their personal data.

19.2

Where available, objections relating to Model Improvement may be submitted through privacy or data settings in the Service, an available privacy request form or another automated means made available by VectorPitch.

19.3

Where no product control, privacy request form or automated means is available, objections should be submitted to privacy@vectorpitch.com.

19.4

If an objection relates to Customer Content controlled by a customer, VectorPitch may need to involve the relevant customer to identify the relevant content and assess the objection.

19.5

The person making the objection may need to provide sufficient information to allow VectorPitch or the relevant customer to identify the relevant content or material.

19.6

If the relevant material has been effectively anonymised so that individuals are not identified or identifiable by means reasonably likely to be used, rights such as erasure and objection may no longer apply to that anonymised material.

19.7

If VectorPitch determines that personal data remains in scope and the objection is valid under Applicable Data Protection Laws, VectorPitch will take appropriate steps. Such steps may include restriction, deletion, further de-identification, exclusion from future Model Improvement datasets or other measures reasonably appropriate to the risk and legal requirement.

19.8

VectorPitch will respond to objections relating to VectorPitch controller processing in accordance with the timing set out in the Privacy Notice.

Section 20 — De-identification, Anonymisation and Challenges

Material that genuinely can't be linked back to a person is treated as anonymous and falls outside GDPR. If you think specific training material wasn't properly anonymised, we'll review that in good faith.

20.1

VectorPitch uses de-identification measures designed to reduce direct identity linkage and privacy risk in Model Improvement datasets.

20.2

De-identification measures may include:

(a) face masking;
(b) audio removal;
(c) metadata severance;
(d) removal or separation of match identifiers;
(e) removal or separation of roster metadata;
(f) removal or separation of organisation and workspace identifiers;
(g) removal or separation of direct player identity metadata;
(h) restriction of access;
(i) use of derived data rather than raw identifiable video; and
(j) separation of operational workspace content from Model Improvement datasets.

20.3

Where material has been processed in a manner that cannot reasonably be used to re-identify any individual, whether by VectorPitch alone or in combination with other information reasonably likely to be used, VectorPitch may treat that material as anonymous data for the purposes of Applicable Data Protection Laws.

20.4

The anonymisation position in clause 20.3 is intended to be assessed in accordance with Recital 26 GDPR and applicable regulatory guidance on anonymisation, including the criteria set out in Opinion 05/2014 of the Article 29 Working Party on anonymisation techniques, as endorsed, replaced or supplemented by the European Data Protection Board or applicable supervisory authorities from time to time.

20.5

VectorPitch will reassess its de-identification measures for Model Improvement datasets at least annually and following any material change in re-identification technology, regulatory guidance, processing context or Model Improvement safeguards that may materially affect re-identification risk.

20.6

If a customer or data subject reasonably challenges whether specific De-identified Training Material has been effectively anonymised, VectorPitch will review the challenge in good faith.

20.7

While a challenge is being reviewed, VectorPitch will apply appropriate protective measures to the disputed material where reasonably practicable. These may include restricting further use, restricting access, reprocessing the material or applying additional de-identification measures.

20.8

VectorPitch will attempt to resolve such challenge within 30 calendar days of receiving sufficient information to identify and assess the disputed material.

20.9

If VectorPitch determines, or is required by a competent supervisory authority or court to treat, specific De-identified Training Material as personal data, VectorPitch will take appropriate remedial steps in accordance with Applicable Data Protection Laws. Such steps may include deletion, further de-identification, restriction of processing, exclusion from future Model Improvement datasets or other measures reasonably appropriate to the risk and legal requirement.

20.10

Nothing in this Notice requires VectorPitch to retrain models already developed using De-identified Training Material unless required by Applicable Data Protection Laws, a competent supervisory authority, a court or the applicable Subscription Plan.

Section 21 — Retention of Model Improvement Material

We keep training datasets while they're genuinely needed and review them at least annually. Deleting footage from your workspace doesn't automatically delete anonymised training material or force retraining.

21.1

Model Improvement datasets may be retained while reasonably needed for model evaluation, training, testing, validation, auditability, quality assurance and improvement, subject to review at least annually under clause 21.2.

21.2

Model Improvement datasets are subject to review at least annually.

21.3

VectorPitch may delete, archive, restrict, further de-identify or replace Model Improvement datasets where they are no longer reasonably needed, where safeguards change, where a customer exclusion applies, where an objection is upheld, where required by law or where VectorPitch otherwise determines deletion or restriction is appropriate.

21.4

Operational Customer Content retention is addressed separately in the Terms, Data Processing Agreement and Privacy Notice.

21.5

Deletion of Customer Content from a customer workspace does not necessarily require deletion of De-identified Training Material already created or retraining of models already developed, unless required by Applicable Data Protection Laws, a competent supervisory authority, a court or the applicable Subscription Plan.

Section 22 — Security and Access Controls

Access controls, tenant separation and restricted internal access protect AI features, outputs and training material. Our people can reach that material only where there's a legitimate reason.

22.1

VectorPitch applies technical and organisational measures designed to protect AI Features, AI Outputs, Customer Content and Model Improvement datasets.

22.2

Such measures may include:

(a) role-based access controls;
(b) logical tenant separation;
(c) restricted internal access;
(d) private cloud storage;
(e) encryption in transit;
(f) encryption at rest;
(g) web application firewall controls;
(h) logging and monitoring;
(i) threat detection and alerting;
(j) vulnerability management;
(k) security testing;
(l) incident response procedures;
(m) restricted access to Model Improvement datasets; and
(n) de-identification measures for Model Improvement datasets.

22.3

VectorPitch personnel may access Customer Content or Model Improvement material only where required for support, troubleshooting, security, abuse investigation, legal compliance, quality investigation, Model Improvement or other purposes described in the Terms, Data Processing Agreement, Privacy Notice or this Notice.

Section 23 — Customer Notices and End-User Transparency

Telling players, parents and others how their data is used is the customer's responsibility.

23.1

Customers are responsible for providing appropriate notices to individuals whose personal data is included in Customer Content.

23.2

Such individuals may include players, under-18 players, parents, guardians, opposition players, coaches, referees, analysts, staff, spectators and incidental individuals appearing in footage.

23.3

Customer notices should explain, where applicable:

(a) that video footage may be uploaded to VectorPitch;
(b) that footage may be analysed using AI Features;
(c) that AI Outputs may be generated;
(d) that AI Outputs may require human review;
(e) whether the customer's plan includes Model Improvement;
(f) whether youth or academy footage is included;
(g) who controls the Customer Content;
(h) how individuals can exercise privacy rights; and
(i) how the customer uses AI Outputs in its own decision-making processes.

23.4

VectorPitch may provide template language or guidance to assist customers with their own notices. Such language is provided for convenience only and does not remove the customer's responsibility to provide legally adequate notices.

Section 24 — Customer Use of AI Outputs

Don't use the AI to discriminate, to watch people unlawfully, to identify anyone by face or body, to read emotions, or to make a serious decision about someone — jobs, selection, welfare — on its own.

24.1

Customers are responsible for all decisions made using or informed by AI Outputs.

24.2

Customers must ensure that AI Outputs are used only by appropriate Authorised Users and for appropriate purposes.

24.3

Customers must not use AI Outputs:

(a) for unlawful discrimination;
(b) for unlawful surveillance;
(c) for biometric identification;
(d) for biometric categorisation;
(e) for emotion recognition;
(f) as a medical or injury assessment;
(g) as a safeguarding decision system;
(h) as an employment decision system;
(i) as the sole basis for selection, scholarship, contract, disciplinary, welfare or eligibility decisions; or
(j) for any High-Risk Use unless expressly agreed in writing with VectorPitch and permitted by applicable law.

24.4

Customers must maintain appropriate human oversight, internal controls and review procedures when using AI Outputs.

24.5

Customers must ensure that their Authorised Users understand the limitations of AI Outputs.

Section 25 — Accuracy, Testing and Performance

We test and monitor how the models perform, using test datasets, de-identified material and your corrections, and we may update models, thresholds and output formats over time.

25.1

VectorPitch may test, evaluate and monitor AI Feature performance.

25.2

Performance may be evaluated using test datasets, customer-derived De-identified Training Material, user corrections, reported errors, product telemetry, quality metrics and other lawful sources.

25.3

VectorPitch may update models, thresholds, confidence scores, workflows, output formats and AI Feature behaviour from time to time.

25.4

Updates may improve, reduce or otherwise change performance for particular footage types, teams, environments or use cases.

25.5

Unless expressly agreed in writing, VectorPitch does not guarantee:

(a) specific accuracy levels;
(b) specific detection rates;
(c) specific false-positive or false-negative rates;
(d) specific processing times;
(e) specific availability of AI Features;
(f) consistency of outputs across different model versions; and
(g) suitability for any particular customer decision.

Section 26 — Documentation and AI Literacy

We provide documentation, warnings, confidence indicators and guidance. Making sure your users are trained enough to understand the limits of what they're relying on is your responsibility.

26.1

VectorPitch may provide documentation, guidance, tooltips, warnings, product labels, confidence indicators, support materials or other information to help customers and Authorised Users understand AI Features.

26.2

Customers are responsible for ensuring that their Authorised Users have an appropriate level of training, context and supervision for their use of AI Features and AI Outputs.

26.3

Customers are responsible for ensuring that users who rely on AI Outputs understand the relevant limitations, intended purpose, need for human review and potential consequences of misuse.

Section 27 — Changes to AI Features and this Notice

We may update the features and this Notice. Material changes come with at least 30 days' advance notice.

27.1

VectorPitch may update AI Features from time to time.

27.2

VectorPitch may update this Notice from time to time.

27.3

Where changes are material, VectorPitch will provide at least 30 days' advance notice to customers or users, for example through the Service, by email or through a website notice, unless a shorter period is required for legal, security, fraud prevention or urgent operational reasons.

27.4

Where a change materially affects Model Improvement, lawful basis, use of Customer Content, third-party AI providers, High-Risk Use classification or customer rights, VectorPitch will provide notice or obtain any required agreement before the relevant change takes effect, where required by Applicable Data Protection Laws or the Terms.

27.5

Subject to clause 27.4, continued use of the Service after an updated Notice takes effect will constitute acknowledgement of the updated Notice, subject to any mandatory rights or requirements under Applicable Data Protection Laws.

Section 28 — Contact

How to reach us about this Notice, about security, and about incorrect AI outputs.

28.1

Questions about this Notice may be sent to:

Vector Sport Solutions Limited
Email: privacy@vectorpitch.com

28.2

Security matters may be sent to security@vectorpitch.com.

28.3

Questions about incorrect AI Outputs may be submitted through available support or product channels.

Schedule A — AI Feature Categories

Every AI feature listed individually — video classification, player detection, ball detection, tracking, event detection and the rest — with its current position on whether output is AI-generated or AI-assisted.

A.1

Video classification. Classification of uploaded video type, such as match footage, training footage, broadcast footage, panoramic footage or highlights. Current position: May be available depending on product configuration.

A.2

Player detection. Detection of players in sports footage. Current position: AI-generated or AI-assisted.

A.3

Ball detection. Detection of the ball in sports footage. Current position: AI-generated or AI-assisted.

A.4

Pitch detection. Detection of pitch features, markings or layout. Current position: AI-generated or AI-assisted.

A.5

Player tracking. Tracking of player positions and movement over time. Current position: AI-generated or AI-assisted.

A.6

Movement metrics. Speed, distance, acceleration, positioning and related sports metrics. Current position: AI-generated or AI-assisted.

A.7

Automated tagging. Automated creation or suggestion of tags and events. Current position: AI-generated or AI-assisted.

A.8

Tactical analysis. Formation, shape, spacing, heatmap, passing, event or tactical outputs. Current position: AI-generated or AI-assisted.

A.9

Confidence scoring. Confidence or quality indicators for certain AI Outputs. Current position: Indicative only, not a guarantee.

A.10

Reports and summaries. AI-generated or AI-assisted reports, summaries or analysis outputs. Current position: May be available depending on product configuration.

A.11

Third-party AI enrichment. AI inference or enrichment using third-party providers. Current position: Optional or plan-dependent where enabled.

A.12

Model Improvement. Evaluation, training, fine-tuning, testing and improvement of AI Features. Current position: Plan-dependent and subject to safeguards.

Schedule B — Model Improvement Safeguards

The specific safeguards applied to training material: no raw identifiable video retained, effective de-identification required before use, face masking, audio removal and metadata severance.

B.1

No raw identifiable video retained in training corpus. Customer Content used for Model Improvement is not retained as raw identifiable video in the training corpus.

B.2

Effective de-identification requirement. Material that cannot be effectively de-identified in accordance with this Notice will not be used for Model Improvement.

B.3

Face masking. Faces are masked in material used for Model Improvement where technically and operationally practicable.

B.4

Audio removal. Audio is removed from material used for Model Improvement.

B.5

Metadata severance. Match, date, organisation, workspace, roster and direct player identity metadata is removed, severed or separated using appropriate technical and organisational measures.

B.6

De-identified Training Material. Model Improvement is designed to use de-identified frames, crops, derived tracking data, tags, corrections, annotations and related outputs.

B.7

Corrections and annotations included in governance review. Corrections, annotations, confirmations, rejections and reported errors are considered in the LIA and DPIA where used for Model Improvement involving personal data.

B.8

No named-person identification training. Customer Content is not used to train models to identify named individuals.

B.9

No biometric templates. Model Improvement is not intended to create biometric templates for uniquely identifying individuals.

B.10

No Special Category Data inference. Model Improvement is not intended to infer health, injury, ethnicity, religion, political opinion, sexual orientation or other Special Category Data.

B.11

No third-party model training. Customer Content is not provided to third-party AI providers to train their models.

B.12

Restricted access. Access to Model Improvement datasets is restricted.

B.13

LIA. Legitimate interests assessment has been or will be conducted where GDPR applies and legitimate interests is relied on.

B.14

DPIA. Data protection impact assessment has been or will be conducted for Model Improvement processing.

B.15

Children considered. The LIA and DPIA consider the possibility of children, opposition players, spectators and incidental individuals in source footage.

B.16

Annual de-identification review. De-identification measures are reassessed at least annually and following material changes affecting re-identification risk.

B.17

Objection handling. Objections and anonymisation challenges are handled through the process described in this Notice.

B.18

Plan-level exclusions. Model Improvement exclusion may be available depending on the Subscription Plan.

B.19

No VectorPitch legal-effect decisions. VectorPitch does not use AI Outputs to make decisions producing legal or similarly significant effects concerning individuals.

Schedule C — Customer Responsibilities for AI Use

A checklist of what sits with you: footage rights, youth consents and safeguarding, human review of outputs, and appropriate access controls.

C.1

Footage rights. Ensure all required rights, permissions, notices, consents and lawful bases exist before upload.

C.2

Youth footage. Ensure appropriate parent, guardian, safeguarding, club, academy, school, league and competition permissions.

C.3

Human review. Review, correct and validate AI Outputs before relying on them.

C.4

Decision-making. Do not use AI Outputs as the sole basis for High-Risk Use.

C.5

User training. Ensure Authorised Users understand AI limitations and intended purpose.

C.6

Notices. Provide appropriate notices to players, staff, parents, guardians and other relevant individuals.

C.7

Access controls. Manage user permissions, roles, teams, squads and external sharing.

C.8

External sharing. Ensure clips, playlists, reports and links are shared lawfully and appropriately.

C.9

AI law compliance. Comply with Applicable AI Laws where the customer's use triggers regulated deployer or user obligations.

C.10

Error reporting. Report incorrect AI Outputs where appropriate.

Schedule D — Lawful Basis and Governance for Model Improvement

The formal basis for model improvement — independent controller, legitimate interests under Article 6(1)(f) — with the interest being accuracy and reliability on real-world sports footage, plus the governance around it.

D.1

Processing role. Independent controller for Model Improvement carried out for VectorPitch's own purposes.

D.2

Primary lawful basis where GDPR applies. Legitimate interests under Article 6(1)(f) GDPR.

D.3

Legitimate interest. Improving sports analysis accuracy, reliability, safety, consistency and performance on real-world customer footage.

D.4

Necessity rationale. Public and professional broadcast datasets do not adequately represent many customer environments.

D.5

Consent assessment. Consent is not the primary basis because many individuals in footage are not direct VectorPitch users and obtaining, recording and managing reliable informed consent from all individuals appearing in sports footage would be impracticable in many real-world match contexts.

D.6

Balancing safeguards. De-identification, no raw identifiable video retained in training corpus, no named-person identification training, no biometric templates, restricted access, objection handling and periodic review.

D.7

Children's data. Youth and academy source footage may be present, and this is specifically considered in the LIA and DPIA.

D.8

Opposition and incidental individuals. Opposition players, referees, spectators, parents, guardians and incidental individuals are specifically considered in the LIA and DPIA.

D.9

Corrections and annotations. Corrections, annotations, confirmations, rejections and reported errors are considered where used for Model Improvement involving personal data.

D.10

LIA. Conducted or to be conducted before relying on legitimate interests for Model Improvement involving personal data.

D.11

DPIA. Conducted or to be conducted for Model Improvement processing.

D.12

Objection process. Available where legitimate interests applies and personal data remains in scope.

D.13

Automated objection route. Objections may be submitted through product settings, a privacy request form or another automated means where available, with email as fallback.

D.14

Anonymised material. Data protection rights may no longer apply where material is effectively anonymised.

D.15

AI Act classification. Launch AI Features have been or will be assessed against Applicable AI Laws based on intended purpose, design, marketing, instructions, restrictions and usage controls.

Schedule E — Model Improvement by Subscription Plan

Whether model improvement is included, excluded or optional on each plan. A signed agreement or order form overrides this Schedule where they conflict.

E.1

Purpose of this Schedule. This Schedule describes whether Model Improvement is included, excluded, optional or subject to customer control under each Subscription Plan.

E.2

Precedence. If there is a conflict between this Schedule and a separate written agreement, order form or signed agreement between VectorPitch and the customer, the separate written agreement, order form or signed agreement will prevail to the extent of that conflict.

E.3

Updates. VectorPitch may update this Schedule from time to time in accordance with this Notice, the Terms of Service Agreement and Applicable Data Protection Laws.

E.4

Model Improvement plan positions. The current Model Improvement position for each Subscription Plan is set out in the table below.

Subscription Plan	Model Improvement Position	Customer Control	Notes
Beta	Included unless VectorPitch confirms otherwise in writing	Exclusion only if confirmed by VectorPitch in writing or made available through a product control	Model Improvement status must be disclosed before first upload under the beta plan
Paid plans	As displayed during plan selection or before first upload	As displayed during plan selection; an exclusion may be available through a product control or written confirmation from VectorPitch	Plan-specific positions will be added to this Schedule when paid plans launch

E.5

Youth, academy and school content — review. Where a customer uses the Service for youth, academy, school or similar content, the customer should review the Model Improvement position before uploading that content.

E.6

Youth, academy and school content — exclusion. If the customer requires youth, academy or school content to be excluded from Model Improvement, the customer must select a Subscription Plan that includes that exclusion, enable the relevant product control if available, or obtain written confirmation from VectorPitch before the relevant content is processed for Model Improvement.

E.7

Prospective effect. Unless otherwise agreed, a Model Improvement exclusion applies prospectively only.

E.8

Effect of exclusion. Unless required by Applicable Data Protection Laws, a competent supervisory authority, a court or the applicable Subscription Plan, a Model Improvement exclusion does not require:

(a) deletion of De-identified Training Material already created;
(b) deletion of anonymised material;
(c) retraining of models already developed; or
(d) deletion of statistical or aggregate model performance information.

E.9

Disclosure before first upload. Where a Subscription Plan includes Model Improvement, VectorPitch will take reasonable steps to disclose that position before the first upload of Customer Content under that Subscription Plan.

E.10

Disclosure records. VectorPitch may record the relevant disclosure, acceptance, plan configuration or upload confirmation for audit, legal, compliance and service administration purposes.