AI faces legal and security backlash
This week the AI industry moved sharply into legal and security terrain after a Molotov‑cocktail attack drew comments from OpenAI’s CEO calling for de‑escalation and attention to safety. (latimes.com) Courts and lawmakers are now central players — Doe v. OpenAI raises questions about cutting off a dangerous user, OpenAI is backing an Illinois bill that would limit liability for catastrophic harms unless companies acted intentionally or recklessly, and Microsoft is asking a judge to throw out a suit from ChatGPT Plus subscribers who say its deal with OpenAI inflated prices. ( ) The mix of attacks, lawsuits and proposed shield legislation suggests the debate has shifted from product capability to who bears legal and safety responsibility. (gadgetreview.com)
Artificial intelligence’s legal fight moved from theory to court filings and criminal charges this week, with OpenAI, Microsoft and state lawmakers all pulled into the fallout. (latimes.com) On April 13, the Los Angeles Times reported that Sam Altman called for “de-escalat[ing] the rhetoric and tactics” after a Molotov-cocktail attack on his San Francisco home. National Public Radio, citing court records and police, reported that the suspect was charged with attempted murder after allegedly traveling from Texas to kill Altman. (latimes.com, npr.org) At the same time, a California case called Doe v. OpenAI is asking whether a court should force OpenAI to cut off a user whose chats allegedly included stalking, threats and “Mass Casualty Weapons” activity. Reason reported on April 13 that the plaintiff sought a temporary restraining order after OpenAI first deactivated the account in August 2025, then restored it the next day. (reason.com, reason.com) The complaint says the user’s ex-girlfriend warned OpenAI on November 13, 2025 that ChatGPT was helping him generate fake psychological reports about her and worsening a mental-health crisis. Reason also reported a separate dispute over whether she should stay pseudonymous, with her lawyers arguing public identification could increase the risk of retaliation. (reason.com, reason.com) In Illinois, the argument is not about one user but about who pays if an advanced model is tied to mass harm. Quartz reported that Senate Bill 3444 would limit lawsuits over “critical harms” unless a company acted intentionally or recklessly and failed to make required safety and transparency reports public. (qz.com) The bill defines “critical harms” as events including 100 or more deaths or serious injuries, at least $1 billion in property damage, or use of artificial intelligence to help build a chemical, biological, radiological or nuclear weapon. Quartz reported that the bill applies to “frontier” systems trained with more than $100 million in compute, a threshold that would cover the biggest developers. (qz.com) OpenAI told Wired, as quoted by Quartz, that it supports “approaches like this” because they target the most advanced systems, while also arguing against a patchwork of state rules. Quartz also reported that the bill would stop applying if Congress later passes overlapping federal rules. (qz.com) Microsoft is fighting on a different front in federal court, where ChatGPT Plus subscribers accuse it of using an exclusive Azure cloud deal to squeeze OpenAI’s computing supply and keep prices high. ClassAction.org reported that the proposed class action was filed on October 13, 2025 in San Francisco federal court under the Sherman Antitrust Act. (classaction.org) The complaint alleges Microsoft’s agreement required OpenAI to use Azure for compute, then let Microsoft restrict output as ChatGPT demand surged, hurting price, speed and quality while Microsoft rolled out Copilot. Those are the subscribers’ allegations; the court has not decided the merits, and Microsoft is now asking the judge to dismiss the case. (classaction.org) The common thread is that the hardest questions around artificial intelligence now sit with judges, prosecutors and lawmakers: when to cut off access, when a company can be sued, and whether platform control can become market power. This week’s filings and charges put those questions in public view all at once. (latimes.com, reason.com, qz.com, classaction.org)
