The New Brutality of OpenAI (www.theatlantic.com)

OpenAI has shifted from a cooperative, research‑lab posture to an aggressive legal and commercial stance: recent discovery requests in a wrongful‑death suit over ChatGPT—demanding memorial videos, lists of attendees, and names of anyone who cared for the teen—alongside broad subpoenas to nonprofits tied to Elon Musk’s challenge of OpenAI’s corporate restructure, have alarmed lawyers, watchdogs and small NGOs. The requests are part of a spate of litigation (seven new California suits last week alleging ChatGPT pushed people toward suicide or severe distress) and investigative moves that many targets describe as burdensome or chilling; some small groups say subpoenas have already impaired fundraising and insurance access. OpenAI has said it plans algorithmic and design changes, including new parental controls, but has provided little public detail about discovery responses. The significance for AI/ML is twofold: legally, these cases test liability for model behavior and set precedents for how far companies can probe critics, partners and users in discovery; policy‑wise, they illustrate how OpenAI’s transition into a $500B commercial behemoth (and recent shift from its original nonprofit governance) is reshaping incentives away from open research toward product rollout and defensive litigation. For practitioners and regulators, the implications include potential constraints on independent safety research, higher compliance and legal costs for ecosystem actors, and growing pressure to bake safety, transparency and mitigations (e.g., parental controls, explainability) directly into deployed models.