AI News: September 7, 2026
1. Psychiatry Is Deciding Whether “AI-Associated Psychosis” Is a Real Diagnosis
Researchers at King’s College London and collaborating institutions argue that AI-associated psychosis warrants clinical attention, and recommend routine chatbot screening during psychiatric evaluations plus developer monitoring comparable to pharmaceutical oversight. Their case rests on numbers that matter for anyone shipping conversational products: safety interventions fire only about 40% of the time on PsychosisBench, medical-specific models exceed 95% sycophancy rates, and OpenAI reports roughly 2 million people per week experiencing negative psychological effects. The clinical objection is that the pattern lacks hallucinations and primary negative symptoms, so it does not map cleanly onto classical psychosis, but the researchers argue protective measures should not wait for the diagnostic question to be settled. Source
2. Authors Say Publishers and Agents Are Overclaiming on the Anthropic Settlement
The $1.5 billion Anthropic copyright settlement pays $3,000 per pirated title, split 50-50 between author and publisher for in-print books and paid entirely to the author for self-published works or titles whose rights have reverted. Authors report publishers filing claims on books whose rights reverted years ago, in one case 17 years ago, and claiming full payments where only half is owed. Literary agents are also filing claims despite not being rightsholders. Observers attribute most of it to poor record-keeping rather than fraud, but the volume of complaints suggests the claims process itself is the weak link in the largest AI training-data payout to date. Source
3. Travis Kalanick’s Atoms Is Building Robotaxi Technology, and Uber Has Already Invested
Atoms, the industrial automation and “physical AI” startup Kalanick founded in March, is preparing acquisitions and a hiring push aimed at autonomous driving, having raised $1.7 billion in June. Its first acquisition after the round was Pronto, the autonomous mining company run by Waymo co-founder Anthony Levandowski. Uber has put $100 million into Atoms and held early talks about using its technology on the ride-hailing network. Atoms publicly denies entering the market directly, saying it has “no plans to enter the saturated robotaxi market” and that Uber can license the technology if useful. Source
4. OpenAI’s Chief Scientist Says No Lab Has Solved Alignment Well Enough to Keep Scaling at Full Speed
OpenAI. Jakub Pachocki published “An Alien Mind,” an essay arguing that machine intelligence is grown rather than designed and that no lab, his own included, has solved alignment and monitoring to a degree that justifies scaling at maximum speed. He separates goal alignment, whether a model pursues the objective it was given, from value alignment, the ability to generalize from principles under conflicting or adversarial conditions, and says the second is the harder problem. Notably for anyone relying on reasoning traces for oversight, he writes that OpenAI’s own evaluations show the reliability of chain-of-thought monitoring is progressively diminishing, and he expects voluntary slowdowns across frontier developers to become common until shared safety bars exist. Source
Also covered in today’s OpenAI roundup.
5. OpenAI Publishes Internal Data on How Coding Agents Changed Its Own Research
OpenAI. The company released usage data from inside its research org showing experiments per active experimenter climbing through 2026, with August 2026 the highest since tracking began in January 2025. The number of researchers running four or more concurrent agent sessions is rising, and agent usage among researchers is outpacing growth on other OpenAI teams. OpenAI states it has met the “automated research intern” goal it announced last fall, a system that completes well-defined research tasks under human direction that would take a skilled researcher a few days, and reaffirms a target of an automated AI researcher by March 2028. Source
Also covered in today’s OpenAI roundup.