Responsible AI Policy

1Purpose and scope

1.1This Responsible AI Policy (the “Policy”) sets out how Mine Marketing LTD, an Israeli company whose registered address is 28 HaArba'a St., Tel Aviv-Yafo 6473925, Israel (“Mine Marketing”), builds, uses and controls artificial intelligence in its services. Its governing principle is that the Agents apply changes inside limits, and people set the limits.

1.2It applies to every AI System that Mine Marketing builds or uses in providing its services, including the Agents and the Third-Party AI Services used within them, on every Client account and every Owned Property, and to the people who configure, supervise and review them. AdsBuzz at ads.buzz has its own policies.

1.3This Policy is read with the Advertising & Brand-Safety Policy and the Privacy Policy. An Agreement may add to this Policy but does not relax it.

2Definitions

2.1In this Policy, “including” means “including without limitation”, and:

Account Lead

The named person at Mine Marketing who is accountable for a Client account and for every Agent action on it.

Agents

The AI agents that Mine Marketing operates on one data layer: the Opportunities Researcher, Creative, Optimization and Scaling agents, and the ad-review agent (the “Ad-Review Agent”).

AI System

Any system that uses a Model to produce forecasts, recommendations, content, decisions or actions, including the Agents.

Change Log

The record of every change to an account described in Section 10.

Client

A business that has signed a written agreement for services with Mine Marketing (an “Agreement”, which includes any document made under it).

Hard Limits

The limits on each account, including budget ceilings, step sizes, Platform policies and brand-safety rules, that are enforced in code outside the Models.

Model

A statistical, machine-learning or language model used by an AI System, including its version, configuration and instructions.

Operating Mode

The mode agreed with a Client for its account (Supervised, Guardrailed or Fully autonomous) that determines which changes the Agents may apply without a person’s prior approval (Section 6).

Owned Property

A website that Mine Marketing operates itself, including one monetised with licensed search and native feeds (search arbitrage).

Partner

A provider of licensed search or native feeds carried on an Owned Property, or another partner through which Mine Marketing monetises traffic.

Personal Data

Any information relating to an identified or identifiable individual, including “personal information” as defined in the data-protection laws that apply.

Platform

An advertising platform on which Mine Marketing buys media, as listed in the Advertising & Brand-Safety Policy.

Site

The website at https://mine-marketing.com, including all of its pages and the policies published on it.

Sign-off

Approval of a proposed change by the Account Lead, or by another person the Account Lead has authorised for the account.

Third-Party AI Service

An AI model or service supplied by another company and used within an AI System.

3Principles

3.1Mine Marketing designs, uses and supervises AI Systems according to these principles:

1. Human oversight and accountability. People set objectives, budgets and Hard Limits, approve what matters and answer for every account.
2. Safety and reliability. AI Systems are tested before use, monitored in use and stopped or rolled back when they misbehave.
3. Transparency. Clients know which Operating Mode applies and what the Agents do, and every change can be traced.
4. Fairness and non-discrimination. AI Systems are not used to target or exclude people unlawfully or unfairly.
5. Privacy and data minimisation. AI Systems use only the data needed, only for the Client it belongs to, and no Personal Data to train Models.
6. Security. AI Systems and their data are protected against unauthorised access, misuse and manipulation.
7. Compliance. AI Systems are used in accordance with the law, Platform policies and each Agreement.

4Governance

4.1A named Account Lead is accountable for every Client account and for every Agent action on it, and oversees the application of this Policy to that account.

4.2Mine Marketing maintains an internal inventory of its AI Systems, recording for each its purpose, the Models and Third-Party AI Services it uses, the data it uses, its owner, its risks and controls, and the date of its last review.

4.3A person reviews every change to a Model, including a new Model, a new version, retraining or a change to its configuration or instructions that affects its outputs, before it replaces the one in use. The review evaluates the new Model against the one in use and is recorded with its result and the approver.

4.4Everyone who configures, supervises or reviews AI Systems is trained on this Policy, on the AI Systems they work with and their limitations, and on the Platform policies relevant to their work, on joining and at least once a year.

5The AI Systems

5.1The Agents work on one shared data layer. Four of them each own one stage of the campaign lifecycle, and the Ad-Review Agent checks creative:

Opportunities Researcher

Finds and forecasts demand before any budget is spent.

Creative

Generates, tests and rotates ad and landing-page creative.

Optimization

Moves bids, budgets and pacing toward return.

Scaling

Expands what works across markets and platforms, step by step.

Ad-Review Agent

Reviews every image and video ad against Platform policy and brand-safety rules, before launch and in rotation.

5.2Statistical and machine-learning Models forecast demand, allocate budget, test creative and detect anomalies. Language Models prepare text for people to review and summarise information. Every proposed change passes the Hard Limits and the policy and brand-safety checks before it is applied, and reaches a Platform only through that Platform’s official API.

6Operating Modes

6.1Each account runs in the Operating Mode agreed with the Client:

Supervised

Every change is approved by a person before it is applied.

Guardrailed (the default)

The Agents act inside the limits; policy-sensitive changes wait for the Account Lead’s Sign-off.

Fully autonomous

Machine-learning optimisation of bids, budgets, pacing and the rotation of approved creative runs inside the Hard Limits, with continuous monitoring and automatic rollback. Anything outside the limits waits for Sign-off.

6.2An account runs in Guardrailed mode unless the Client agrees another mode in writing. A change of mode is agreed with the Client, signed off by the Account Lead and recorded in the Change Log, with the date it takes effect. A Client may move any account to a more supervised mode at any time under Section 17.

7Hard Limits

7.1In every Operating Mode, the Hard Limits are enforced in code outside the Models, so that no Model can override, change or disable them. A proposed change that would breach a Hard Limit is not applied. Where the limit is one that a person may change, such as a budget ceiling or step size, the change waits for Sign-off and the limit is changed only by a person; a change that would breach a Platform policy, a brand-safety rule or the law is rejected.

7.2The Hard Limits and the objectives and budgets of each account are set and changed by people. No Agent sets or changes them.

8Human approvals

8.1In every Operating Mode, a person approves:

1. new opportunities, before budget is spent on them;
2. entry into a new market or onto a new Platform;
3. budget ceilings and any change to them; and
4. every new ad and every new landing page, by a final review before it goes live.

8.2No Agent publishes a new ad, enters a new market or Platform, or raises a budget ceiling on its own authority.

9Monitoring, anomaly holds and rollback

9.1Mine Marketing monitors the Agents’ actions and account performance continuously. Statistical Models detect anomalies, such as unexpected spend, sudden changes in cost or conversion, or a breach of pacing.

9.2When an anomaly is detected, the Agents hold further automated changes on the affected campaign, the Account Lead is alerted, and the hold remains until a person has reviewed it. In Fully autonomous mode, a change that degrades results beyond the limits set for the account is rolled back automatically to its previous value, where the Platform allows.

10Logging and audit

10.1Every change to an account, whether made by a person or an Agent, is logged with its reason, the limit checked, the approver and the previous value, so that it can be reviewed and, where the Platform allows, rolled back.

10.2The Change Log for a Client’s accounts is available to that Client and is provided on request within 10 business days. Each Platform’s own change history gives the Client a record that is independent of Mine Marketing’s.

10.3The Change Log, Model review records and AI Incident records for an account are kept for the duration of the Agreement and 24 months after it ends, unless the Agreement or the law requires otherwise.

11Data

11.1The data the Agents use for an account is set by the Client’s Agreement and may include campaign, conversion and Platform-reporting data, and market and search-demand data that Mine Marketing is entitled to use. The Agents obtain Platform data only through official APIs.

11.2Client data is used only for that Client. One Client’s data is not used for, disclosed to or used to improve the service of another Client.

11.3No Personal Data is used to train, fine-tune or evaluate Models. A Model that learns from campaign performance uses only aggregated, non-personal data of the Client concerned.

11.4AI Systems receive only the data needed for their task. Where Mine Marketing processes Personal Data for a Client, it does so as the Client’s processor or service provider, on the Client’s instructions and under the Agreement. Retention follows the Privacy Policy and, for Client data, the Agreement.

11.5Mine Marketing does not use the Agents to make decisions about individuals that produce legal effects concerning them or similarly significantly affect them, within the meaning of Art. 22 of the GDPR.

12Third-Party AI Services

12.1Mine Marketing uses a Third-Party AI Service only under terms that keep Client data confidential, that do not permit the provider to use it to train or improve the provider’s models, and that meet the data-protection and transfer requirements described in the Privacy Policy.

12.2Each Third-Party AI Service is assessed before use for security, data handling, reliability and legal compliance, recorded in the inventory, and reassessed at least once a year and when its terms change.

13Security

13.1Mine Marketing protects its AI Systems, their data and the Platform access they use with measures appropriate to the risk, including:

1. access to AI Systems, Models and Platform credentials limited to authorised people who need it, with individual accounts and multi-factor authentication where the system supports it;
2. Platform access granted to the Agents with only the permissions their task requires, and withdrawn when it is no longer needed; and
3. treating content from outside an account, such as web pages, documents and messages, as data and not as instructions to an AI System, and testing AI Systems against manipulation of their inputs.

14Generative content safeguards

14.1Content produced by AI Systems, including ad copy, images, video and landing pages, passes the policy and brand-safety checks and the Ad-Review Agent where applicable, and a person gives it a final review before it goes live.

14.2AI-generated or altered content is disclosed and labelled where Platform rules or the law require. Mine Marketing does not produce deceptive synthetic media, does not impersonate any person or organisation, and does not use a real person’s likeness or voice without their documented permission.

14.3Before use, generative content is checked so that it does not reproduce or imitate another’s protected work or trade mark, and third-party elements are licensed. Further rules are in the Advertising & Brand-Safety Policy.

15Fairness and non-discrimination

15.1The Agents do not target or exclude people on the basis of a protected characteristic, or a proxy for one, in a way that is unlawful or discriminatory. They apply the Platforms’ special ad category rules, including those for credit and financial products, employment and housing, as Hard Limits.

15.2Mine Marketing monitors the Agents’ targeting and delivery for patterns that suggest unfair or unlawful outcomes, and corrects the configuration, data or Model concerned when it finds one.

16Transparency

16.1Each Client is told, before work begins and whenever it changes, which Operating Mode applies, what the Agents do on its accounts and which changes wait for Sign-off. The Site and this Policy describe how the Agents work.

16.2Where an AI System interacts directly with a person, the person is told that they are dealing with an AI System, unless that is obvious from the context.

16.3AI Systems can make mistakes, and forecasts and recommendations are estimates. Nothing in this Policy guarantees any result; the Disclaimer and each Agreement govern what Mine Marketing undertakes.

16.4Questions about the AI Systems from Clients or from anyone affected by them are answered by writing to team@mine-marketing.com with “Responsible AI” in the subject line. Rights over Personal Data are exercised as the Privacy Policy describes.

17Human review, takeover and change of mode

17.1A Client may, at any time, by writing to its Account Lead or to team@mine-marketing.com:

1. ask for a person to review any Agent action, which Mine Marketing completes within two business days and, where the action was wrong, corrects;
2. pause the Agents on any account or campaign, or take over any part of the account, so that changes are made only by people or by the Client; or
3. change the Operating Mode of any account.

17.2A pause, takeover or move to a more supervised mode is applied as soon as practicable and in any event within one business day. A move to a less supervised mode takes effect when agreed and signed off under Section 6.

18Incidents

18.1Where an AI System acts outside its limits, produces harmful or non-compliant output, or behaves in a way that may harm a Client or anyone else (an “AI Incident”), Mine Marketing will:

1. pause the AI System on the affected accounts and roll back affected changes where the Platform allows;
2. notify the affected Client within one business day of confirming the AI Incident, and any Platform, Partner or authority where the law or its policy requires;
3. identify the root cause and correct the AI System, its Model or its limits before it resumes, with a person’s review; and
4. record the AI Incident, the actions taken and the outcome.

18.2An AI Incident involving Personal Data is also handled under the breach procedure in the Privacy Policy and the Agreement. AI Incident records are kept for the period in Section 10.

19Legal compliance

19.1Mine Marketing uses AI Systems in accordance with the data-protection laws described in the Privacy Policy, advertising and consumer-protection law, and the policies of each Platform and Partner.

19.2Where the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) applies, Mine Marketing complies with the obligations that apply to it as provider or deployer, including AI literacy and the transparency duties for AI-generated or manipulated content. It does not use AI Systems for any practice that the Act prohibits, and does not use them for a purpose the Act classifies as high-risk unless the Act’s requirements for it have first been met.

20Review and contact

20.1Mine Marketing reviews this Policy at least every 12 months and whenever there is a material change to its AI Systems, its services or the law, and publishes the current version on this page with its last-updated date.

20.2Questions about this Policy should be sent to team@mine-marketing.com with “Responsible AI” in the subject line, or by post to Mine Marketing LTD, 28 HaArba'a St., Tel Aviv-Yafo 6473925, Israel.