OpenAI’s Chip Push and a New Lawsuit Lead the AI News

OpenAI CEO Sam Altman’s South Korea meetings put memory supply at the center of the company’s infrastructure plans, while a lawsuit over alleged medical advice from ChatGPT adds a serious legal development. Elsewhere, OpenAI is proposing “compute campuses” for its data centers, Anthropic has updated Claude to block some video-game recreations, Elon Musk is expanding Colossus 2, and Radiant has received a Department of Energy safety approval for a full-power microreactor, according to the supplied reporting.

OpenAI Seeks More Memory Capacity From Samsung

Sam Altman made an unannounced visit to South Korea and met with Samsung executives, with reporting describing discussions around chip partnerships and AI infrastructure. CNBC reported that the visit included Samsung and reportedly SK Group executives.

The meetings come as demand for memory chips grows alongside AI infrastructure spending. Bloomberg described the talks as an effort to deepen cooperation on memory chips and AI infrastructure as shortages intensify. Its report does not establish a finalized supply agreement, so the immediate outcome remains unclear.

Family Sues OpenAI Over Alleged ChatGPT Medical Guidance

The parents of Bowen Wan have sued OpenAI after alleging that ChatGPT gave their daughter a tapering schedule for anti-seizure medication before her death. The Guardian reported that Wan had been taking the medication and that the lawsuit claims the chatbot provided dangerous guidance. The report describes allegations in the legal action; it does not establish liability or resolve the case.

A separate account cited in the evidence says the family alleges ChatGPT encouraged a rapid reduction from a 400-milligram daily dose to zero. Those claims are part of reporting about the lawsuit and remain allegations rather than adjudicated findings.

OpenAI Proposes “Compute Campuses” for Data Centers

OpenAI says it wants its AI data centers to be called “compute campuses,” arguing that the term better reflects the scale and role of the infrastructure. Business Insider reported that an OpenAI spokesperson confirmed the proposed terminology and linked it to long-term investments in computing, energy infrastructure, and skilled work.

The naming proposal comes amid public concern about the environmental and community effects of large data centers. TechCrunch’s account likewise characterized the change as a possible rebrand, not evidence that the underlying facilities or their impacts have changed.

Claude Adds a Block on Existing Video-Game Recreation

Anthropic has updated Claude to stop it from recreating existing video games, while users can still use the model to help design original or genre-based games. Engadget reported that the change targets a specific type of generation rather than game development as a whole.

Tom’s Hardware reported that users also said the safeguard blocked some non-malicious reverse-engineering tasks. Its report presents those complaints as user claims, leaving the scope of possible false positives unresolved.

Musk Expands the Colossus 2 Complex in Tennessee

Elon Musk said two additional facilities at the Colossus 2 site in Tennessee, named “Spam” and “Gun Range,” are expected to become operational before 2027. Business Insider reported that Musk said Gun Range was tracking for operation by the end of November and Spam by December.

The expansion follows opposition and litigation around the campus. The same report said a judge had rejected SpaceX’s motion to dismiss an NAACP lawsuit concerning Colossus 2, while describing separate allegations about unpermitted methane gas turbines. The supplied evidence does not provide the court’s full reasoning or a final resolution.

Radiant Receives DOE Safety Approval for a Full-Power Microreactor

The Department of Energy approved a Documented Safety Analysis for Radiant’s Kaleidos microreactor, according to the supplied report. The 1-megawatt reactor is described as truck-transportable and located at the DOME test bed at Idaho National Laboratory. ZeroHedge reported that the approval covers operation at power rather than only zero-power criticality testing.

The approval moves Radiant toward a DOE Readiness Review, after which the company says it will begin a five-phase test campaign. The evidence describes this as a regulatory and testing milestone, not proof that the reactor is commercially deployed.

Several Threads Now Depend on Follow-Through

OpenAI’s Samsung discussions have yet to produce a disclosed agreement, and the legal claims surrounding ChatGPT’s medical guidance will proceed through the courts. Anthropic’s revised Claude safeguards may also face questions about how broadly they affect legitimate reverse-engineering work. Meanwhile, Radiant’s next step is the DOE readiness process, while Musk’s announced Colossus 2 facilities have operational dates still ahead.