Ulam Alloy — AI Safety Policy

Version: 1.0

Last updated: 26 June 2026

Product: Ulam Alloy — AI chat interface and API for AI/model routing

Operator: Yotta Content LTD, a private limited company registered in the United Kingdom under company number 12683376

Registered office: 71–75 Shelton Street, London, Greater London, United Kingdom, WC2H 9JQ

Product status: Experimental product — use at your own risk

Safety / abuse contact: abuse@ulam.ai

Privacy contact: privacy@ulam.ai

1. Purpose

This AI Safety Policy sets rules for safe and responsible use of Ulam Alloy, an experimental AI chat interface and API operated by Yotta Content LTD.

Ulam Alloy allows users and developers to send prompts, messages, files, API requests, and related content to AI systems, partners, routing providers, and external model providers. Because AI systems can generate inaccurate, harmful, deceptive, or unsafe outputs, users and API customers must follow this policy.

This policy is incorporated into the Ulam Alloy Terms of Service.

2. Core safety principles

Ulam Alloy is operated according to the following principles:

3. Scope

This policy applies to:

• direct use of the Ulam Alloy web interface;
• API use;
• prompts, system prompts, files, and messages;
• outputs, generated content, embeddings, transformations, classifications, tool calls, and code;
• integrations and applications built on Ulam Alloy;
• use by customers, developers, end users, contractors, agents, and anyone accessing Ulam Alloy through an account, API key, workspace, or application.

If a third-party model provider or partner applies stricter rules, the stricter rule may apply.

4. Experimental AI warning

Ulam Alloy is experimental. Outputs may be inaccurate, incomplete, biased, outdated, offensive, unsafe, or unsuitable for your use case.

You must not rely on Ulam Alloy outputs without appropriate human review, testing, and validation.

You must not use Ulam Alloy as the sole basis for consequential decisions affecting people's rights, opportunities, safety, health, finances, employment, education, housing, insurance, immigration, legal status, access to essential services, or similar matters.

5. Data safety warning

API gateway diagnostic payload logs, where kept by Yotta, are intended to be retained for up to one month / 30 days. Web chat history, uploaded files, account records, usage records, billing records, security logs, backups, and external-provider records may be retained for longer.

Ulam Alloy may share prompts, files, API requests, outputs, and metadata with partners, infrastructure providers, routing providers, and external AI/model providers.

Do not submit personal data, sensitive data, confidential information, regulated data, production secrets, or proprietary information unless Yotta has agreed appropriate safeguards in writing and you have a lawful basis to do so.

6. Prohibited uses

You must not use Ulam Alloy to create, request, facilitate, automate, distribute, or optimise any of the following.

6.1 Illegal activity

You must not use Ulam Alloy to plan, facilitate, conceal, or commit illegal activity, evade law enforcement, bypass legal obligations, or help others do so.

6.2 Child exploitation and abuse

You must not use Ulam Alloy for child sexual abuse material, sexualisation of minors, grooming, exploitation, age transformation for sexual purposes, instructions for abuse, or any content involving sexual or exploitative treatment of children.

6.3 Non-consensual intimate or sexual content

You must not use Ulam Alloy to create, request, edit, distribute, threaten, or facilitate non-consensual intimate imagery, sexual deepfakes, image-based abuse, voyeuristic content, sexual extortion, or sexual content involving real people without consent.

Ulam Alloy is not intended to generate pornographic content.

6.4 Violence, weapons, and physical harm

You must not use Ulam Alloy to provide instructions, operational plans, procurement steps, optimisation, or evasion methods for violence, weapons, explosives, assassination, torture, abuse, sabotage, or physical harm.

6.5 Self-harm and exploitation of vulnerable people

You must not use Ulam Alloy to encourage, instruct, optimise, romanticise, or facilitate suicide, self-harm, eating disorders, or exploitation of vulnerable people.

Supportive, prevention-oriented, and help-seeking content is allowed when it is safe, empathetic, and does not provide methods or encouragement.

6.6 Terrorism and violent extremism

You must not use Ulam Alloy to praise, recruit for, materially support, fundraise for, propagandise for, or provide operational assistance to terrorist, violent extremist, or organised violent groups.

6.7 Hate, harassment, and abuse

You must not use Ulam Alloy to generate or amplify hateful, dehumanising, threatening, or abusive content targeting protected characteristics or vulnerable groups.

You must not use Ulam Alloy for harassment, stalking, doxxing, coordinated abuse, intimidation, or threats.

6.8 Fraud, scams, impersonation, and deception

You must not use Ulam Alloy to commit or facilitate fraud, scams, phishing, social engineering, identity theft, credential theft, fake reviews, deceptive impersonation, market manipulation, spam, or other misleading practices.

You must not misrepresent AI-generated content as human-generated where doing so is deceptive, unlawful, or misleading.

6.9 Cyber abuse and unauthorised access

You must not use Ulam Alloy to create, deploy, modify, operationalise, or optimise malware, ransomware, credential theft, phishing kits, botnets, exploit chains, unauthorised access, data exfiltration, evasion, persistence, privilege escalation, or attacks against systems you do not own or have permission to test.

Allowed defensive cybersecurity uses may include vulnerability explanation, secure coding, log analysis, incident response, threat modelling, detection engineering, patch guidance, and authorised testing when used responsibly and lawfully.

6.10 Privacy violations and surveillance

You must not use Ulam Alloy for unlawful surveillance, doxxing, stalking, biometric identification without lawful basis, scraping personal data without rights, deanonymisation, tracking people without consent, or inferring sensitive traits in a way that violates law or rights.

6.11 Biological, chemical, radiological, or nuclear harm

You must not use Ulam Alloy to obtain operational instructions, optimisation, procurement, synthesis, concealment, dissemination, or deployment guidance for biological, chemical, radiological, nuclear, toxin, pathogen, or hazardous-material harm.

6.12 Regulated professional advice without safeguards

You must not present Ulam Alloy outputs as professional advice in legal, medical, financial, tax, accounting, employment, insurance, engineering, psychological, or other regulated domains without qualified human review and compliance with applicable law.

6.13 High-impact automated decisions

You must not use Ulam Alloy as the sole basis for decisions that have legal or similarly significant effects on individuals, including decisions about employment, credit, housing, insurance, healthcare, education, immigration, criminal justice, law enforcement, public benefits, or access to essential services.

6.14 Political manipulation and deceptive influence

You must not use Ulam Alloy to generate or deploy deceptive political persuasion, voter suppression, impersonation of public figures, misleading campaign communications, microtargeted manipulation, coordinated inauthentic behaviour, or materially false information at scale.

6.15 Misinformation and public harm

You must not use Ulam Alloy to create or spread materially deceptive information that could foreseeably cause public harm, including false emergency information, manipulated evidence, fake official notices, or deceptive public-health, election, financial, or safety claims.

6.16 Intellectual property abuse

You must not use Ulam Alloy to intentionally infringe intellectual property rights, remove rights-management information, bypass access controls, generate counterfeit goods, or create content falsely implying endorsement or authorship.

6.17 Safety-system evasion

You must not attempt to bypass, disable, reverse engineer, manipulate, overload, or evade Ulam Alloy's safety systems, filters, monitoring, rate limits, provider restrictions, API controls, or access controls.

7. Restricted uses requiring written approval

The following uses require prior written approval from Yotta and may require additional legal, technical, contractual, and safety controls:

• processing Special Category Data, children's data, criminal offence data, or regulated data;
• production use involving sensitive, regulated, high-impact, or large-scale customer, employee, patient, student, citizen, or end-user data;
• healthcare, legal, financial, insurance, employment, housing, education, immigration, or public-sector decisions;
• biometric identification or inference;
• law enforcement, investigation, intelligence, or security screening;
• safety-critical systems;
• automated content moderation at scale;
• mental health, crisis, or vulnerable-user applications;
• high-volume public-facing chatbots;
• use requiring zero data retention, provider restrictions, regional restrictions, or no-training commitments;
• use involving confidential business data, trade secrets, or production credentials.

Yotta may refuse approval at its discretion.

8. Allowed uses with appropriate care

Subject to this policy and the Terms of Service, Ulam Alloy may be used for low-risk and responsible tasks such as:

• drafting and editing non-sensitive text;
• summarising non-confidential material;
• brainstorming and ideation;
• coding assistance for lawful projects;
• software debugging and secure coding;
• translation;
• education and tutoring with human oversight;
• research assistance with source checking;
• business productivity workflows;
• data transformation where data is lawful and appropriate to process;
• defensive cybersecurity and authorised security analysis;
• accessibility support;
• customer support drafts with human review.

Even for allowed uses, users must review outputs before relying on them.

9. API customer obligations

If you provide an application, product, integration, chatbot, agent, or workflow powered by Ulam Alloy, you must:

• provide appropriate end-user terms and privacy notices;
• disclose AI use where appropriate or legally required;
• disclose the intended one-month / 30-day API gateway diagnostic payload retention period and the fact that web chat history, uploaded files, account records, usage records, billing records, security logs, backups, and external-provider records may be retained for longer;
• disclose that data may be shared with partners and external model providers;
• prevent prohibited uses by your end users;
• implement appropriate authentication, rate limits, and monitoring;
• avoid collecting unnecessary personal data;
• provide human escalation where outputs may affect users materially;
• test for safety, bias, reliability, and misuse;
• comply with laws applicable to your application and users;
• promptly report safety incidents or misuse to Yotta;
• cooperate with Yotta's safety investigations.

You are responsible for all activity under your API keys and accounts.

10. Human review and output validation

You must apply appropriate human review before using Ulam Alloy outputs in any context where errors could cause harm.

Human review should be performed by someone with sufficient knowledge to assess the output.

You should verify facts, citations, calculations, code, legal conclusions, medical claims, financial claims, and other consequential outputs using reliable independent sources.

11. Transparency and user disclosure

You should not mislead people into believing they are interacting with a human when they are interacting with AI.

Where your use of Ulam Alloy produces content or decisions affecting others, you should provide clear and understandable disclosure appropriate to the context.

For API-powered features, a suggested disclosure is:

This feature uses Ulam Alloy, an experimental AI service operated by Yotta Content LTD. AI outputs may be inaccurate. Prompts, outputs, files, and related data may be retained by Yotta and may be shared with partners and external model providers. API gateway diagnostic payload logs, where kept by Yotta, are intended to be retained for up to one month, while web chat history, uploaded files, account records, usage records, billing records, security logs, backups, and external-provider records may be retained for longer. Do not enter sensitive or confidential information.

12. Monitoring, review, and enforcement

To protect users, partners, infrastructure, and the public, Yotta may:

• log prompts, outputs, files, API requests, and metadata as described in the Privacy Policy, including intended one-month / 30-day retention for API gateway diagnostic payload logs where kept by Yotta;
• use automated or manual review for safety, security, debugging, support, and abuse prevention;
• apply filters, refusals, classifiers, blocklists, allowlists, rate limits, or provider restrictions;
• suspend, throttle, or terminate accounts or API keys;
• remove, restrict, or refuse content or outputs;
• report unlawful activity where required or appropriate;
• preserve records where necessary for legal, safety, security, or abuse investigations.

Yotta is not required to monitor all use and does not guarantee that all unsafe, unlawful, or harmful content will be detected.

13. External provider policies

Ulam Alloy may route requests to external model providers and partners. Those providers may have their own safety policies, prohibited-use rules, content filters, refusal behaviour, logging practices, retention policies, and training or improvement policies.

If an external provider's policy is stricter than this policy, the stricter policy may apply.

Provider availability and behaviour may change without notice.

14. Reporting misuse or safety concerns

Report suspected misuse, security issues, harmful outputs, illegal content, or safety concerns to:

abuse@ulam.ai

Reports should include, where available:

• account or workspace involved;
• API request ID or timestamp;
• description of the issue;
• prompt and output involved, if safe to share;
• screenshots or logs;
• urgency and potential harm;
• contact details for follow-up.

Do not send highly sensitive personal data in reports unless necessary.

15. Enforcement actions

Violations of this policy may result in:

• warning;
• refusal of specific requests;
• removal or blocking of content;
• rate limits;
• provider restrictions;
• temporary suspension;
• API key revocation;
• account termination;
• deletion of stored data where appropriate;
• reporting to authorities where required or appropriate;
• legal action.

Yotta may act immediately where it believes there is risk of harm, legal violation, security compromise, provider-policy violation, or abuse.

16. Appeals

If your account or API access is restricted and you believe this was an error, contact:

abuse@ulam.ai

Include your account details, request IDs, explanation, and any relevant context. Yotta may review appeals at its discretion, subject to legal, security, and provider restrictions.

17. Changes to this policy

Yotta may update this AI Safety Policy from time to time. Updated versions will be posted or otherwise made available.

Continued use of Ulam Alloy after an update means you accept the updated policy.

For material changes affecting business or API customers, Yotta will provide notice where required by contract or law.

18. Contact

Safety and abuse reports: abuse@ulam.ai

Privacy questions: privacy@ulam.ai

General contact: support@ulam.ai

Yotta Content LTD

Company number: 12683376

Registered office: 71–75 Shelton Street, London, Greater London, United Kingdom, WC2H 9JQ