105 stories
External testing found Claude could reject a harmful request yet miss a deceptive campaign split into ordinary tasks. The midterm update also details voting-information plans and remaining political-bias problems.
In a newly published account of his September discussion, Whyte connects unexpected AI actions with studies showing that perceived human involvement changes confidence in identical information.
The proposal would examine agent behavior, monitoring and AI-assisted research. It seeks public evidence, not a safety certification, and leaves developers substantial control over what assessors can see.
Some developer changes remain promises. Meanwhile, ongoing enquiries into agent testing will examine safeguards, with public evidence due by November 20.
SB 1050 covers realistic fictional figures and generated voices. Its prominence test is narrower than New York’s, but it also assigns removal duties to ad distributors.
FCC chair threatens E-Rate fund, blames schools for kids' excessive screen time.
Plaintiff claims cameras enabled warrantless tracking and monitoring after he previously sued over excessive force.
OpenAI engineer faces 5 years and $250K fine as FBI lobs foreign spy charges.
“We should look for love!” Transportation Secretary Sean Duffy bellowed while flanked by his nine children at the Great American State Fair in June: “Get married, and have lots of kids!” In a brazen attempt to make Duffy’s ethos a matter of policy, the Trump administration announced a strange new…
Attack method exposes flaws in OpenAI, Google safeguards against AI nudes.
The recommendations could inform state legislation. A public draft may arrive in weeks to months, with feedback promised before adoption.
Their bipartisan letter calls for measurable standards, independent model testing and wider access to defensive tools. It asks the administration to act while Congress develops legislation.
The court cited a risk of evidence being destroyed and approved early depositions. Apple’s allegations remain contested, and several research requests must wait.
Disney may acquire tech from one of the first chatbot makers linked to child suicide.
The October 16 session will give lawmakers a chance to question the company as they weigh local AI protections. The subpoena allows a representative to appear instead of Elon Musk.
A watchdog alleges the companies conceal who actually gets paid. Leading the Future denies wrongdoing, while the election regulator lacks the votes needed to decide complaints.
The investigation questions both accuracy checks and promised time savings. Applicants cannot inspect or correct the AI summaries of their interviews.
AB 1979 takes effect January 1, 2027. It preserves AI-assisted care while requiring health facilities to protect professional judgment.
Showing 18 of 105 stories
Load more storiesGet the top AI news and tools straight to your inbox every week.
Weekly · AI news
Join 15,000+ AI enthusiasts
