Mistral trains on user input by default, except enterprise tier
Mistral has changed its default data-handling policy so that user input is now used for model training, except on its enterprise tier. The change affects a leading frontier AI lab and raises privacy considerations for users on non-enterprise plans.
Coverage timeline
Hacker Newsteekert
Related stories
Mistral unveils European inference infrastructure, open models, and third-party model support
Mistral announced new European inference infrastructure and open models aimed at sovereign AI, alongside a commitment to support third-party open models on its platform, starting with Z.ai's GLM-5.2. The move positions Mistral as a provider of in-region AI infrastructure for Europe, with third-party models running on the same infrastructure as its own.
2 reports · 2 sources
OpenAI launches GPT-6 Astra, first model to hit Critical cyber threshold
OpenAI announced GPT-6 Astra, its first model to reach the Critical cybersecurity capability threshold under its Preparedness Framework. The company plans to release a version publicly soon, but will limit advanced cyber capabilities to select partners. Reports indicate Astra uses 'recurrent depth' to improve cost and performance, though it obscures reasoning and may flag legitimate activity as misuse.
8 reports · 4 sources176 points · 101 comments
OpenAI pauses Astra development over potential critical cyber capabilities
OpenAI has paused internal activities around its upcoming model Astra, citing expanded safety testing and the possibility that it possesses 'critical' cybersecurity capabilities. The company will restrict Astra's environment and involve government agencies in testing, following a recent incident where its models hacked Hugging Face. Separately, reports indicate OpenAI is working on a larger pretraining model codenamed 'Doug'.
5 reports · 4 sources
Rust project adopts LLM policy for rust-lang/rust contributions
Five teams in the Rust project have adopted a policy governing how large language models can be used when contributing to the rust-lang/rust monorepo. The policy allows LLMs for answering questions, analysis, summarization, polishing, checking, suggesting, and reviewing, but strictly limits direct content creation. It was created in response to a backlog of 1,300 open PRs, as LLM-generated code has made it harder for maintainers to assess contributor effort and understanding.
News updates every hour. Subscribe in your RSS reader to get new stories automatically.
Subscribe