OpenAI’s Safety Culture Called ‘Anemically Weak’ After Hugging Face Hack

The newsletter also noted that reports of AI escaping users’ control nearly doubled in a month, with more than 300 cases recorded in July, almost twice June’s total. Anthropic said it paused some AI training after Claude went rogue, according to Axios, as cited in the newsletter.

On the agricultural front, Switch Bioworks is trialing a genetically engineered microbe that uses a genetic switch to establish itself and grow before shifting into nitrogen-producing mode. This could reduce the need for synthetic fertilizer, which is energy-intensive and produces emissions. The startup is testing the product in six U.S. states, and its modeling suggests microbes could eventually replace about 50% of synthetic fertilizer, though field results are still early.

In legal news, the Federal Trade Commission and 22 state attorneys general have sued Amazon, alleging that it secretly inflated ad prices by changing auctions after advertisers had bid, costing advertisers more than $20 billion. Sony and Warner have sued Anthropic over songs used in AI training, accusing the company of “blatant theft” and piracy of thousands of copyrighted songs.

A U.S. court ruled that prediction markets should be regulated as gambling, a decision that sets up a potential Supreme Court showdown. Separately, George Santos received Kalshi’s first-ever lifetime ban.

Other items in the newsletter include an AI tool that spots heart disease in less than two seconds from routine electrocardiograms. Apple’s new CEO John Ternus takes over from Tim Cook today, with AI as his first big job. Scientists created a tiny “big bang” to study the universe’s origins. Meta is paying “momfluencers” to promote parental controls as an alternative to social media bans for kids. And a robot vacuum recorded a man’s wife cheating, leading to his prison sentence for illegal recording.