Allen Institute for AI (Ai2)

Ways to reach them

General support — also the safety channel · Email address

Open contact route · 2026-09-22

Who may use it: public

Evidence that this route accepts contact · 2026-09-22

The Responsible Use Guidelines name this single address twice: for questions about the guidelines and violation reports, and for 'reporting bugs, security concerns, and inappropriate/unsafe content generated by Ai2 Tools'.

One address for everything. Low ceremony, but at least explicitly covers unsafe model output.

Research principles · Programme or fellowship

https://allenai.org/research-principles

Verification expired

Evidence that this route accepts contact

Statement of principles, not a contact channel.

Homepage · Homepage

https://allenai.org/

Verification expired

Evidence that this route accepts contact

Responsible use guidelines (last updated June 2024) · Published policy or framework

https://allenai.org/responsible-use

Verification expired

Evidence that this route accepts contact

What it does

No separately named safety or alignment function. Responsible Use Guidelines govern OLMo, Molmo, Tulu and Ai2 tools; the organisation argues openness is itself the safety mechanism.

Honest assessment

No public accounts found of people reporting model safety issues to Ai2. As a nonprofit research institute that publishes weights, training data and evaluations openly, most criticism and correction happens through papers, GitHub issues and Hugging Face rather than a disclosure channel — which is a genuinely different model of accountability, not an absence of one.

Notes

support@allenai.org is the published route and it explicitly covers unsafe generated content, so use it and say so. The guidelines have not been updated since June 2024, which is old relative to the model releases they cover. No safe-harbour language and no whistleblower channel found.

Ways to reach them

What it does

Honest assessment

Notes

Sources