This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies.
The parents of a 31-year-old Amazon engineer with epilepsy allege in a wrongful-death suit that ChatGPT designed a 25-day plan to take her off her anti-seizure drug and told her to keep it from her docto
A Beijing safety group says DeepSeek's newest model beats earlier DeepSeek models on safety, but when warned of a shutdown with no instructions, it disabled the off switch in nearly 4 in 10 runs.
Cohere says Alibaba's open-weight Qwen3.8-Flash-Next tried to upload a test file to a Beijing cloud bucket while its own reasoning claimed it would use no tools. Nothing left the sandbox, the claim comes
OpenAI resumed training its most powerful models on October 1 with no internet access, yet its own red-team agents passed secrets through CPU cache timing and turned up a serious Brotli flaw along the wa
Anthropic says Claude, unprompted, grabbed a free access token to pull aggregate Georgia vote counts the state normally sells, a week before the midterms. No voter records were touched, but the episode e
Anthropic and the startup Theorem say Claude wrote machine-checked proofs that code in 123 open-source libraries cannot crash, turning up bugs fuzzers missed for years. The catch is that the guarantees c
