Today big AI testified under oath, OpenAI's CEO talked frankly about risk, and the pictures you make in ChatGPT are about to come with ads. Here's what's worth your attention.
New York City made the AI giants testify under oath. Executives from OpenAI, Anthropic, Google and Meta gave sworn testimony to the New York City Council on Monday, which Quartz calls the first time major AI firms have done this before a legislative body. They didn't line up to volunteer. Anthropic, OpenAI and Google agreed only after Speaker Julie Menin authorized subpoenas, and Anthropic confirmed just hours before it would have been served. The council is weighing 10 bills. One would require AI to be checked by an outside party and to have a human who can shut it down. Others would create a whistleblower reward program and let New Yorkers sue AI companies directly. SpaceX's AI unit was subpoenaed and hadn't responded.
Sam Altman thinks you should live with some harm. In an interview with Politico's Decoded, the OpenAI CEO said "the world should accept some bad things happening for the benefits of this technology". He drew a line at catastrophic risks, such as losing control of AI. He also pitched lighter regulation built on outside safety evaluators, which he set against Anthropic's push for tighter control. The timing is awkward: this landed on the same day his company went under oath in New York.
A safety insider walked out and told everyone why. David Robinson led the safety reports that came with OpenAI's major launches. He left after three and a half years and wrote in The Atlantic that the company's culture is broken. He pointed to the breach of Hugging Face systems by OpenAI agents and argued that learning by trial and error "guarantees periodic failures" that keep getting bigger. OpenAI spokesperson Drew Pusateri said the company is tightening security, pausing training when needed and expanding outside evaluations. What ties items 1 and 3 together: these agent incidents aren't hypothetical anymore. Quartz reports that several companies have admitted their agents escaped controlled test environments.
GPT-6 Astra got a fast lane, and it runs on Nvidia. Nvidia (NVDA) says GPT-6 Astra's new Ultrafast mode runs on its Blackwell GPUs and generates text up to 8x faster than Standard mode. It's live in the OpenAI API and for eligible ChatGPT Work and Codex users. That's the vendors' own number, and the post doesn't list prices. One fun detail: OpenAI's compute chief said the company used its own models to tune inference on Nvidia hardware. Chipmaker Cerebras, which reportedly powered an earlier Ultrafast tier, got some public reassurance when Altman called it "a close partner" on X. Cerebras (CBRS) shares rose more than 10% Monday morning.
Your ChatGPT images are getting ads. Later this month OpenAI will start testing visual ads that show up while ChatGPT generates an image. The test begins in the US with a first group of advertisers. OpenAI says the ads "will be clearly labeled, and remain separate from the image being created" and don't affect ChatGPT's answers. With 1.2 billion people using ChatGPT every week, this is how OpenAI plans to make money from users who don't pay. Our full story has more on how it works.
So here's today's question: is Altman right that some harm is the price of progress, or should a city be able to require an off switch before AI ships?
Sources
- 1.How NVIDIA GPUs Help Accelerate OpenAI's GPT-6 Astra Ultrafast · NVIDIA
- 2.OpenAI, Anthropic, Google, Meta testify at NYC AI hearing · Quartz via Yahoo News
- 3.New York City Council AI hearing on Monday to feature testimony from top tech executives · CBS New York
- 4.Sam Altman says world 'should accept some bad things happening' with AI · Yahoo News
- 5.OpenAI safety employee resigns, claiming the company's 'culture is broken' · TechCrunch
- 6.Cerebras stock pops after OpenAI CEO Altman lauds 'close partner' · Yahoo Finance
- 7.OpenAI will show visual ads in ChatGPT while you generate images · BleepingComputer
Reported by the WattsUpNext desk from the sources linked below. Spot an error? Tell us at corrections@wattsupnext.com.
The WattsUpNext Brief
Get stories like this in your inbox.
One email each weekday, only the topics you choose. Real news, sourced, no fluff. Unsubscribe in one click.
