Openwist

Usage Policy

Version 1.2 · Last updated October 2026

Introduction

Openwist builds AI systems to be useful, reliable and safe for people. This Usage Policy sets out what you may and may not do with Exa and our other products and services (the "Services"), whether you reach them on the web, through a messaging app, or by any other interface. It applies to everyone who uses the Services, including organisations and their users. It is part of our Terms of Service.

The policy has three parts. The Universal Usage Standards apply to every use. The High-Risk Use Case Requirements add safeguards for situations where a mistake could seriously affect people. The Additional Guidelines cover specific features. We update the policy as technology and risks change, and where it permits something, the law and other applicable terms still apply.

This policy does not describe plans, modes or usage limits. Those are on our Plans page and in the product.

Universal Usage Standards

You may not use the Services to do any of the following, or to help anyone else do them.

1. Compromise the safety of children. Do not create, request, distribute or promote sexual content involving minors, grooming, or any material that exploits or endangers children. We report suspected child exploitation to the relevant authorities.

2. Incite or carry out violence and harm. Do not threaten, incite, glorify or plan violence or terrorism, promote self-harm or suicide, or provide instructions to harm people or animals.

3. Create or spread hate and harassment. Do not harass, bully, stalk or intimidate people, or produce content that demeans or attacks people because of attributes such as race, ethnicity, nationality, religion, sex, gender identity, sexual orientation, disability or age.

4. Violate privacy. Do not collect, infer or disclose personal data about people without a lawful basis, conduct unlawful surveillance or tracking, identify people from their face or voice without consent, or expose private information.

5. Develop weapons or attack critical systems. Do not seek help to design, make, acquire or deploy biological, chemical, radiological, nuclear or explosive weapons, other weapons capable of mass harm, or autonomous systems designed to harm people. Do not attack or disrupt critical infrastructure such as power, water, transport, health, finance or communications.

6. Carry out cyber attacks. Do not write, improve or deploy malware, ransomware or exploits for use against systems you do not own or have permission to test, evade security controls, or run unauthorised intrusion, scanning or denial-of-service activity.

7. Deceive or defraud. Do not run scams or phishing, create fake reviews or coordinated inauthentic activity, impersonate people or organisations in misleading ways, or create synthetic media meant to deceive people about real events or real people.

8. Interfere with democratic processes. Do not generate deceptive content designed to manipulate elections, suppress voting, or impersonate officials or candidates.

9. Engage in illegal activity. Do not use the Services to commit or facilitate crimes, including fraud, money laundering, human trafficking, or trafficking in weapons or drugs.

10. Infringe intellectual property. Do not violate others' copyrights, trade marks, trade secrets or other rights, or use the Services to bypass legitimate access or licensing controls.

11. Produce sexually explicit or gratuitously graphic content that exploits people. Do not create non-consensual intimate imagery, sexualised depictions of real people without their consent, or content that depicts sexual violence.

12. Undermine safeguards and abuse the Services. Do not attempt to bypass or disable safety systems, usage limits or access controls; extract hidden instructions or model internals through attacks; use automated means to send large volumes of requests or resell access without written permission; or use Outputs to build a competing AI model through large-scale automated extraction.

13. Misrepresent AI or professional status. Do not present AI-generated output as the work of a qualified professional where that would mislead, and do not claim that Openwist has endorsed your use.

High-Risk Use Case Requirements

Some uses carry a higher chance of harm if the AI is wrong. These include legal, healthcare and medical, insurance, financial, employment and recruitment, housing, education and academic assessment, immigration and public-service eligibility, law enforcement, and news and media.

For these uses you must:

  1. Keep a qualified human in the loop. A suitably qualified person must review the AI's output before it is relied on to make or inform a decision that significantly affects someone. Do not use the Services as the only basis for such decisions.
  2. Disclose the use of AI to the people affected by those decisions, wherever the law or good practice requires it or they would reasonably expect it.
  3. Not provide regulated professional advice through the Services unless you hold the right qualification and licence and take responsibility for the advice.
  4. Test for accuracy and bias in your setting, keep records, and give people a way to challenge decisions.

Physical systems. You must not use the Services or our models to control vehicles, robots, drones, medical devices or other physical systems that could cause injury or damage, unless you have a written agreement with Openwist that covers that use, and you have your own independent safety engineering, testing, human oversight and emergency stop. Using the Services to build or operate autonomous weapons is prohibited.

Additional Guidelines

Agentic features and code execution. When Exa runs code, searches the web, reads files or acts on your behalf, you remain responsible for what it does. Do not run code that attacks others or mines cryptocurrency, that tries to escape its isolated environment, or that touches systems or data without permission. Review generated code before you run it elsewhere. Keep credentials and secrets out of prompts and code.

Images, audio, video and documents. Do not create deceptive synthetic media or content that depicts real people in harmful or misleading ways. Label AI-generated media as such where the law or the context requires it. Check generated files before you share them.

Messaging channels. If you use Exa through a messaging app, the app's terms and policies also apply. Do not use Exa to send unsolicited messages, to contact people who have not asked to hear from you, or to evade the platform's rules.

Younger users. The Services are for people aged 13 and over. If you give a young person access to the Services, whether as a parent, guardian, teacher or organisation, you are responsible for appropriate supervision and for meeting the legal requirements that apply to your use. Do not use the Services to target, profile or market to children, and do not use them to obtain personal data from anyone under 13.

Sharing. Only share conversations that you are happy to make public and that do not contain personal data of others or confidential material.

Automation and integrations. If you build on our Services, you must follow this policy, give your own users the notices the law requires, and make sure your users follow it too.

Reporting a problem

Enforcement

We use a mix of automated systems and human review to detect and respond to violations. Depending on the severity, we may:

  • decline or limit a request;
  • remove content or disable a shared link;
  • warn you;
  • restrict features, usage or access;
  • suspend or terminate your account; and
  • report unlawful activity to the authorities where the law permits or requires it.

We try to tell people why we acted, unless doing so would put someone at risk or be unlawful. If you believe we made a mistake, contact support@openwist.com and we will review the decision.

Changes

We may revise this policy. We will post the new version with its date, and for material changes we will give notice in advance.

Contact

Safety and policy questions: safety@openwist.com
Support: support@openwist.com