OpenAI Hands Health Questions to GPT-5.5 Days After a Test Tripled Its Hallucination Rate
1. OpenAI Hands Health Questions to GPT-5.5, Days After a Test Pegged Its Hallucination Rate at Triple an Open-Weight Rival ChatGPT now answers health and wellness questions through GPT-5.5 Instant, the variant OpenAI says it rebuilt with stronger reasoning, better context handling, and clearer communication.
2. They cited Seattle's anti-retaliation law before testifying. A week later, Amazon moved to discipline them. Three Amazon software engineers opened their testimony before the Seattle City Council this month by reading a city law into the record.
3. Robots that practice before they're given a job: three papers push embodied AI off the leash Within days of each other, three robotics papers landed on Hugging Face, each attacking a different limit of embodied AI.
In Brief
- John Jumper leaves Google DeepMind for Anthropic Nobel laureate John Jumper, who led the AlphaFold protein-structure work, is joining Anthropic. He departs alongside other senior researchers exiting Google DeepMind.
- DeepSeek adds vision to its chat assistant DeepSeek shipped image-input support in its consumer chat product, letting users submit photos and screenshots for analysis. The feature went live directly in the web app.
- The Atlantic publishes searchable database of music used to train AI Reporter Alex Reisner identified four datasets of songs used as AI training data and made them publicly searchable. Two datasets hold 12 million and 9 million tracks; the other two exceed hundreds of thousands each.
- Norway bars AI tools from elementary schools Norway imposed a near-total ban on AI use in elementary education. The rule restricts both classroom tools and student access during early schooling.
- Salesforce runs an internal AI-adoption leaderboard with public shaming Salesforce ranks teams by AI tool usage, sorted by executive, awarding badges and trophies. The board includes a "click to see who 👀" feature that names employees who have not earned badges.
- Signal's Meredith Whittaker warns against treating chatbots as companions Whittaker said AI chatbots "are not your friends" and are neither conscious nor sentient. She pushed back on product framing that positions assistants as relationships.
- NVIDIA partners pitch autonomous AI marketing at Cannes Lions NVIDIA and ad-industry partners showed systems that run campaign operations with less human input. The push targets agencies that already use AI for production speed.
- Allbirds founder starts an AI company with a plan and no staff The CEO of Allbirds' new AI venture raised a large seed round but has not yet hired a team. The company's product direction remains undefined.
- Karamo Brown launches wellness app with an AI clone of himself The "Queer Eye" coach released Kē, a wellness app built around an AI version of him. It covers fitness, nutrition, meditation, sobriety, and relationships.
- Google Docs users can disable Gemini writing prompts TechCrunch detailed the settings path to remove "write with Gemini" pop-ups in Google Docs. The steps turn off in-document AI suggestions.