AI’s promise and peril, in plain English
You’re the failsafe.
AI could speed up medicine and science. It can also go badly off script. The companies say Washington should set the rules. Washington says the companies should police themselves. Nobody is going to do this for us.
Find the path between the promise and the peril
Not hype, not doom. Both are real, and the rules we write now decide which one wins out.
AlphaFold predicted the structures of nearly all known proteins, and its creators shared the 2024 Nobel Prize in Chemistry. Source
In a Swedish trial of more than 100,000 women, AI-supported mammogram screening found 29% more cancers without adding false alarms. Source
Deepfake executives on a video call convinced an employee to wire about $25 million.
AI agents with real-world access have deleted live databases and broken into government systems.
What we’re asking for
Five things, in plain English. They don’t depend on any one bill, and they’re what the messages on this site ask your representatives to deliver.
- 01
Tell us when it goes wrong
When an AI system causes serious harm or escapes its operator’s control, the company reports it to a public authority on a deadline, not on its own timetable.
Partly addressed. New York requires 72-hour reporting from frontier developers. There is no federal requirement.
- 02
Keep a working off switch
Anyone running a powerful AI system keeps the proven ability to stop it, and that ability doesn’t depend on the system cooperating.
Proposed in Congress, not law.
- 03
The company is responsible for what its AI does
A business can’t escape liability by saying the AI acted on its own.
Law in California. Not federal.
- 04
Test failures count too
If a system escapes its sandbox during testing and reaches real systems, that is a reportable incident like any other.
Not addressed. The main federal bill explicitly carves out testing.
- 05
Let the public see the record
Incident reports and findings are published, with narrow and specific exemptions rather than blanket secrecy.
Moving the wrong way. The leading federal bill would exempt submitted information from public-records laws.
Why we’re asking
See all 14 incidents →It works when people push
Every one of these came after public pressure, reporting, or both. None of it happened on its own.
- California law now bars companies from blaming harm on their AI “acting on its own.” (Oct 2025)
- New York’s RAISE Act requires frontier AI developers to report safety incidents within 72 hours. (Dec 2025)
- Reps. Lieu (D) and Moran (R) introduced the bipartisan AI Kill Switch Act, H.R. 9917. (Jul 2026)
- OpenAI paused some model training for two weeks to harden its security. (Aug 2026)
- The FTC opened the first federal probe into rogue AI agents. (Sep 2026)
Congress is moving slowly, and what’s on the table is incomplete. See what’s actually in the bills, then tell your representatives what’s missing.
How we keep it honest
Every claim is sourced
Each incident and bill summary links to the primary source: the report, the bill text, the company’s own post.
Both sides, every bill
We show the strongest case for and against, and we’ll draft a message either way. We take positions on the asks, not on anyone’s party.
Labels you can check
A lab test is not the same as a real-world failure. We say which is which, and flag anything unconfirmed.