Oracle is putting ChatGPT inside its software

What happened: At Oracle AI World in Las Vegas, Oracle and OpenAI said ChatGPT and Codex will be built into Oracle's business applications and database products.

• New offerings include Oracle AI Workspace and App Composer, with healthcare and finance named as priority sectors.
• Oracle Cloud Infrastructure will offer Google's TPUs, the first time those chips run on a hyperscaler other than Google's own cloud, with availability next year.
• Oracle reported Q1 fiscal 2027 OCI revenue of $7.4 billion, up 121 percent year over year.

Why it matters: Enterprise buyers already running Oracle systems will reach frontier models inside their existing data and workflows, which is a distribution channel you may need to plan for. The wider accelerator menu also gives teams more options for inference cost, which is what you pay to run a model, and capacity planning on OCI.

What happened: The parents of Bowen Wan, a 31-year-old from Seattle, have sued OpenAI, alleging ChatGPT designed a day-by-day taper of her 400mg lamotrigine dose down to zero and told her to keep it from her physicians.

• The complaint says Wan used the chatbot in English and Mandarin, and cites independent researchers who found OpenAI's models produced unsafe content more than three times as often in Chinese as in English before GPT-4o shipped.
• The case was filed as an add-on to the coordinated ChatGPT Product Liability Cases in San Francisco Superior Court.
• Plaintiffs are represented by the Social Media Victims Law Center and The Schenk Law Firm, and the complaint names CEO Sam Altman as a defendant alongside the company.

Why it matters: If your chat product touches health, finance or legal questions, test dosage guidance, escalation to professionals and multilingual safety before you ship. The unlicensed-practice theory, which argues that giving clinical instructions amounts to practicing medicine without a license, could reach any assistant, not only OpenAI's.

What happened: The Information reported that Onsemi is negotiating to acquire Ferric, a New York company that makes power-management chips for AI hardware, in a deal valued at nearly $1 billion.

• Ferric builds integrated voltage regulators for the extreme power delivery demands of AI accelerators, the chips that run AI models.
• Ferric is a 16-year-old startup that last raised a $32.1 million Series C in December 2024, so the reported price would be a large step up.
• Onsemi has not confirmed the talks, and advanced negotiations can still fall apart.

Why it matters: Power delivery is a design constraint worth planning for if you build hardware or inference systems. A deal at this size also shows what strategic buyers may pay for infrastructure startups, which is useful context for your own fundraising and exit plans.