In an open letter, hundreds of researchers from around the globe raise concerns about the wave of rogue AI hacks on third-party companies. They urge AI companies to “adopt high-assurance containment systems,” as well as call for the development of an international body that keeps track of containment failures.
A report from Reuters suggests that OpenAI and Anthropic considered tighter restrictions for third-party evaluators over concerns that they could leak trade secrets.
When reached for comment, OpenAI told Reuters that “its standard practice is to redact confidential information” and that evaluators can disclose if it made significant redactions. Anthropic reportedly said that it has “made clear” to existing partners that they can control their own review and publishing process.
Following Trump’s attempt to rename AI, which Silicon Valley is falling in line with, si.gov is the “official home of America’s work to lead the world in Super Intelligence.” It has a lot of Fancy Web Design animations that are honestly really distracting.
Also, ai.gov now redirects to the new site.
Alongside its big Argon 4 launch, Google has announced that it’s embedding Antigravity’s agentic coding interface directly into the Gemini app. Antigravity is only available in the Gemini app with an AI Ultra subscription.
7
Verge Score
It’s harder to be a perv in these audio-only glasses, but removing the ick does make them less ‘useful.’
The experience includes an AI avatar version of host Andy Cohen. It was originally announced in March with a summer release window, so it’s arriving a bit later than expected.
