Google is bringing Gemini Personal Intelligence to Android Auto, letting drivers ask about plans and get answers using information from connected Google apps. The optional feature is starting to reach users in many countries, with cars that have Google built-in set to follow later.
SpaceX said its AI business beat investor expectations in the third quarter, bringing in $7.5 billion in revenue and $431 million in profit. The company is spending heavily on more data centers even as it returned to an overall quarterly profit.
OpenAI has acquired Orchestra Bio, a startup that makes AI tools for managing drug discovery work. OpenAI plans to bring its team and technology into Rosalind, a system for life sciences research, and a related tool for planning and analyzing experiments.
Deepseek has made the software behind its V4.1-Flash model available for others to use and modify. Called DeepInfer, it helps the model answer requests quickly across multiple chips and could lower the cost of running it.
Security researchers found that shared Meta Muse conversations could be found through searchable DNS records, putting private information at risk. They identified 6,800 shared links and found business records, financial details and other sensitive material in some of them.
Claude designed small proteins meant to attach to scorpion neurotoxins, and lab tests found that 17 of 211 designs met the researchers’ binding criteria. The team has not shown that they stop the toxins from causing harm, so they are not an antivenom.
Spotify says it removed 50 million tracks it flagged as spam or low-quality audio from recommendations in 2026 and changed its systems to favor human artists. It has not said those tracks were deleted from the service, and users still cannot filter out all AI-made audio.
ChatGPT is rolling out a feature that lets users lock individual conversations with a PIN or, on supported devices, face or fingerprint recognition. Locked chats are kept apart from regular chat history, but the lock does not change how OpenAI stores or uses their contents.
A bug exposed internal directions for a Meta support chatbot in the code of about 80 WhatsApp help pages. The directions told the bot to avoid legal advice and certain discussions of privacy rules; Meta says it removed the text and fixed the bug.