Security and trust
How Runner protects your data, who processes it, and how to reach us when something looks wrong.
- How is our data protected?
- Kept apart from every other company's, and encrypted in transit and at rest.
- See how
- Whose key protects our passwords and credentials?
- Your organisation's own, in Google Cloud's key service. Closing the organisation destroys it.
- See Pouch
- Is there a data processing agreement?
- Yes. It is published, with SHMAPLEX as your processor.
- Read the DPA
- Is there a third-party report?
- No. Runner has no SOC 2 report and no independent penetration test report.
- Send your questionnaire
Questions about your data
How is each company's data kept apart?
Every company's data is separated in the database itself by row-level security, apart from the identity graph described in the terms of service.
Is our data encrypted?
All traffic is HTTPS, and a browser that has seen the site once refuses plain HTTP afterwards. Supabase, which hosts Runner's database, states that customer data is encrypted at rest with AES-256. strict-transport-security: max-age=63072000; includeSubDomains
How are connected credentials kept?
Credentials you connect are stored encrypted, and the secret parts are never returned to a browser or written to a log. Revoking a connection stops the access.
How do people sign in?
People sign in with an emailed link, a password or a passkey. Organisation owners and admins must also use an authenticator code, or a passkey they confirm on their device.
Who at Runner can read our content?
A few Runner staff can reach what is stored in a workspace, so they can run the service and fix it when it breaks. They do not look at it unless you ask us to for support, it is needed to investigate a security problem or abuse, or the law requires it.
How are the app and its code protected?
The app enforces a content security policy, refuses to be framed, and switches off browser features it does not use. Dependency alerts and security updates are on, and secret scanning with push protection stops a credential reaching the code. content-security-policy: base-uri 'self'; object-src 'none'; frame-ancestors 'none'; form-action 'self'
Is the database backed up?
The release path takes a full backup of the production database before it changes the schema.
What do AI providers receive?
An AI task sends its provider the content it needs, including any personal details typed into that content. Runner does not train AI models on your content. Some providers' own terms let them train on what they are sent, and the providers page says which.
Where does our data live?
Runner's database is in Seoul. Hosting and providers also work in the United States and other countries, and there is no option to keep data in one region.
How long is data kept, and how do we get it out?
Conversation content is deleted 90 days after a conversation closes. An owner or admin can download a takeout from Settings. Closing an organisation stops its sending at once, and 30 days later its data is deleted from Runner's database.
Anyone can check the response headers with curl -sSI https://app.runtheworld.ai. Last checked against production on 1 November 2026.
Report a problem
Email security@runtheworld.ai with what you found, how to reproduce it, and what it lets someone do. You will get a human reply. There is no paid bug-bounty programme.
Please leave out real personal data, credentials, tokens and full session transcripts. A redacted excerpt, or a reproduction on your own test account, is enough.
If you make a good-faith effort to follow this, we will not pursue legal action against you for the research. Test only against accounts you own, do not access or retain anyone else's data, and give us a reasonable chance to fix the issue before making it public.
What to expect after you report
| Stage | Target |
|---|---|
| Acknowledgement of a new report | 2 business days |
| Initial triage and severity assignment | 5 business days |
| Remediation — critical severity | 7 days from confirmation |
| Remediation — high severity | 30 days from confirmation |
| Remediation — moderate severity | 90 days from confirmation |
How personal data is handled is in the privacy statement, and what you agree to is in the terms of service.
