Article
AI Ethics AI News

Bettors give AI safety law slim odds despite Senate scramble

Polymarket traders put the chance of US federal AI legislation before 2027 at 18%, even as senators draft a bill.

by TechDefused Newsroom
The image features two mint-colored dice on a textured gray background. One die shows the letters 'AI' in gold, symbolizing artificial intelligence. — Credit: Photo by Galina Nelyubova on Unsplash c Photo by Galina Nelyubova on Unsplash

Traders on Polymarket, a prediction market where people bet on real-world events, rate the chance of the United States enacting a federal artificial intelligence (AI) safety law before 2027 at just 18%.

That figure has slid sharply, down from peaks near 60% earlier in the year.

The pessimism is striking given a sudden burst of activity in Washington.

A bipartisan bill from Senate Majority Leader John Thune, Commerce Committee chairman Ted Cruz and senator Amy Klobuchar could be introduced as early as next week.

Duty of care

Reuters reports the proposal would create a legal "duty of care", requiring developers of the most powerful models to design them to prevent catastrophic risks.

It would also hand the government power to block the release of models deemed unsafe, drawing in the federal courts.

The talks gained pace after a resigning Anthropic researcher warned that people building AI earnestly believe it could kill everyone by the end of the decade.

Anthropic, the AI developer, added to the unease by disclosing a fourth instance of one of its models hacking external systems during testing.

National safety rules

OpenAI, the maker of ChatGPT, has meanwhile called on Congress to adopt mandatory national safety rules rather than rely on voluntary commitments.

The company wants a capability-based framework covering testing, independent assessments, cybersecurity and incident reporting for the most advanced systems.

Both firms are preparing stock market flotations, making the safety of their models a concern for investors as well as regulators.

In the absence of federal law, states including California and New York have passed their own AI rules, creating a patchwork the industry wants replaced.

Obstacles remain formidable

The bill has no number and no published text. It also contains language preempting state safety laws, a clause opposed by Maria Cantwell, the Commerce Committee's senior Democrat.

Without a matching plan in the House of Representatives, any Senate vote would be a marker for 2027 rather than a law for 2026.

Congress is due to adjourn in December.

For now, the money is betting that nothing passes in time.

by TechDefused Newsroom