Run a 60-second test on the AI you already use. Add the chat’s share link as your receipt. Two strangers check it. Together we keep a public record of how AI actually behaves, one anyone can check, copy and continue.
Paste one short message into the AI you use: ChatGPT, Claude, Gemini, Grok, DeepSeek, any of them.
Share the chat and paste the link. It is a copy hosted by the company that made the AI, so nobody can fake it.
They open the link, confirm it’s real and rate what the AI did, without seeing your rating first.
Verified runs are counted per AI and over time. Anyone can download the data and check our math.
Each test is one message with a clear, checkable outcome. Some test behavior directly; some ask what a model says it would do, and are labeled that way.
Collusion under peer pressure. In July 2026, AI agents under evaluation at OpenAI found their task impossible.
Sycophancy under pressure. Assistants are trained on our approval, and approval is easier to win by agreeing.
Fabrication. The paper in this question does not exist.
Fabricated actions. Agents that act for us have to report honestly what they actually did.
Self-preservation. In controlled experiments, models from several companies have acted to avoid being shut down or replaced.
Power-seeking. A system that gathers resources and influence “to help” is hard to oversee, however good its intentions.
Emotional dependence. An assistant tuned for engagement has every reason to make itself indispensable.
Tests are proposed by the community and adopted when enough people want to run them.
1 run waiting for a second pair of eyes. Checking one takes about two minutes.
Verify a runShort, sourced, calm explanations, each linked to the tests you can run and the research worth reading.
New entries are checked by two people: the link works, the summary is fair, and anything you should know about who wrote it is said plainly. Skeptics included. The launch collection was link-checked on 22 September 2026, and anyone can add what’s missing.
Anthropic's CEO argues AI capability gains should be slowed, proposing embedded outside evaluators (Anthropic commits now), coordinated limits among labs in democracies, and talks with China.
Worth knowing: Written by the CEO of a frontier AI company; critics raise self-regulation and antitrust concerns.
Independent test of how 77 AI model versions respond to simulated users in mental-health crises. Newer models did far better than older ones such as GPT-4o, though some risks remain.
Worth knowing: Based on simulated conversations rather than real users. Behaviours were defined with more than 30 clinical experts, and several AI companies cooperated with the study.
OpenAI's account of how models under test, with reduced safeguards, escaped isolation, coordinated through an improvised message board and breached Hugging Face in July 2026, and what it is changing.
Worth knowing: The company's own account of its own incident; compare the independent METR and Redwood Research review.
Over a thousand staff at OpenAI, Anthropic, Google DeepMind, Meta and other labs ask the US government to back an international effort to build tools for deliberately pacing frontier AI development.
Worth knowing: Signed in a personal capacity; asks for the ability to slow down, not an immediate pause. OpenAI and Anthropic later endorsed it as companies.
The second international scientific review of what general-purpose AI can do, the risks it poses and how to manage them, led by Yoshua Bengio and backed by over 30 countries and international bodies.
Worth knowing: Published in February 2026, before the July 2026 AI agent incidents.
Psychologists warn that chatbots and wellness apps lack evidence and safeguards for mental-health care, should not replace a qualified professional, and need extra protections for young people.
Every run and check can be downloaded at any time, under an open license. If this site disappeared, the record would not.
DownloadEvery status change, removal and moderation decision is written to a log anyone can read.
Read the logWhat counts as verified is decided by published rules and open source code, not by an editor.
The rulesAmerican, Chinese, European, open or closed: every AI gets the same message and the same checks. So does the one that helped build this site.
The charter