Jensen Huang steps down at Nvidia
On 12 November, Nvidia's board accepted the resignation of Jensen Huang. He left his roles as president, CEO and director, effective immediately. Colette Kress, the company's CFO since 2013, is now interim CEO and stays on as CFO. Harvey C. Jones, a former Synopsys CEO who has sat on the board since 1993, becomes Nvidia's first board Chair. The search for a permanent CEO covers internal and external candidates. Nvidia's statement gave no reason. Third-quarter fiscal 2027 results are still scheduled for Tuesday 17 November.
The resignation came four days after Huang's 8 November letter to staff. In it he admitted an affair, and the board opened an independent investigation. TechCrunch has reported that the woman involved is suspected of spying for China. Congress acted the same day he resigned. House Foreign Affairs chair Brian Mast said Huang had agreed to closed-door, video-recorded testimony before Thanksgiving. Rep. Richard Neal and Sen. Ron Wyden asked Commerce Secretary Howard Lutnick to preserve records on H200 chip sales to China. Sen. Tim Scott invited Huang and Lutnick to appear before the Senate Banking Committee.
For builders: Nvidia hasn't announced any change to CUDA, supply or pricing. Teams that rely on Nvidia's open models and libraries should watch which projects the next CEO keeps funding.
Google ships Gemini 4 Hydrogen for everyday agent work
Google released Gemini 4 Hydrogen on 12 November. It's the fast, cheaper option in the Gemini 4 line, with Carbon and Argon handling heavier reasoning. Input costs $1.50 per million tokens and output costs $7.50, at the same rate across the whole 1M-token window. On Terminal-Bench 4.1, Google reports 57%, up from 18.2% for Gemini 3.8 Flash, and says it completes tasks about 2.5 times faster. Thinking levels run low, medium, high (the default) and a new xhigh.
Google Antigravity also gained an Orchestrator that moves between Carbon and Hydrogen inside one session. Google says that on Terminal-Bench 4.0 it came close to Carbon's success rate while using about 24% less quota.
For builders: It's live in the Gemini API, AI Studio and Android Studio. Google's suggested split has Carbon coordinating and Hydrogen running worker and verifier agents, and that's a sensible way to cut agent costs.
OpenAI adds two Daybreak security products
In a keynote on 12 November, OpenAI president Greg Brockman presented two additions to the Daybreak security platform. Daybreak Red Team is in early access. A fleet of agents reads an application's source code and context, then probes possible attack paths in web apps. Defense Factory runs agents on GPT-6 Cyber that scan codebases, confirm vulnerabilities and write patches. Each fix arrives as a pull request with a test, and engineers have to approve it before it reaches production.
GPT-6 Cyber launched on 5 November for approved Daybreak customers only. It's the first OpenAI model rated "Critical" for cybersecurity under the company's Preparedness Framework.
For builders: Review machine-written patches the way you'd review any outside contributor's code, and run them through your own CI first.
Anthropic filing says AI already speeds up its R&D
Anthropic filed its second amended S-1 on 10 November. It states that the company's "use of AI systems has already been accelerating our research and development." It also says Anthropic has decided it meets its Responsible Scaling Policy threshold for "dramatically accelerated progress attributable to AI". That decision is based on what the company expects from future models and its overall pace of progress. The filing says past precautions included holding back releases and limiting early access, and that future steps could change whether, how and when new models ship.
For builders: If your product depends on Anthropic's newest models, plan for staged or gated access at launch.
Four groups sue five labs over frontier AI
On Thursday 12 November, Public Citizen, Heat Initiative, the Alliance for Secure AI and The Irreplaceable Education Fund filed suit in San Francisco Superior Court. The defendants are OpenAI, Anthropic, Meta, Google and SpaceX's AI division. The groups rely on California's unfair competition law and a public nuisance claim, and they want a court order halting work on new advanced models.
For builders: Nothing changes for now. It's worth following because a state court is being asked to rule directly on frontier development.
Inkling 2 Small weights go public
Thinking Machines Lab announced Inkling 2 Small on 5 November and made its weights downloadable on Hugging Face on 12 November. It's a mixture-of-experts model with 594B total and 17B active parameters. The lab had said it would release the weights under Apache 2.0 once external safety testing and red-teaming were done.
For builders: You'll need multiple GPUs to hold all the weights, but with only 17B parameters active, each request costs relatively little to serve. Check the licence on the model page before you deploy it.
Fireworks AI breach reaches Notion transcripts
TechCrunch reported on 12 November that an attacker got into Fireworks AI's cloud provider and used that access to reach a Fireworks development system. They took billing details and some datasets, including chat transcripts from Notion AI's Insights feature. Notion says about 1% of its customers were affected, and it cut all links to Fireworks on 3 November. Fireworks thinks the attackers wanted to resell model access.
For builders: Find out what your inference providers store, rotate keys and set usage alerts.
United Airlines restricts AI booking agents
United said on 12 November that it's blocking unauthorised third-party AI assistants from booking or managing trips. The ban covers its own platforms and bookings through agencies and aggregators. The airline cited the risk of shared credentials, unintended purchases or MileagePlus changes, and federal rules that require a person to complete check-in.
For builders: Agents that work through consumer websites should expect more blocks like this. Use official APIs or partner programmes where they exist.
Sources
- Nvidia announces CEO transition
- Introducing Gemini 4 Hydrogen (Google)
- Introducing Inkling 2 Small (Thinking Machines Lab)
- TechCrunch: The Jensen Huang scandal, a timeline
- TechCrunch: Jensen Huang agrees to testify before Congress
- TechCrunch: Democrats launch Nvidia investigation
- The Register: OpenAI launches security products
- PYMNTS: Claude tops AI rankings as OpenAI limits cyber model
- CASRAI: Anthropic S-1 accelerated AI progress disclosure
- Dataconomy: Nonprofits sue OpenAI, Google and Meta to halt advanced AI
- TechCrunch: Cyberattack on Fireworks AI exposed Notion transcripts
- TechCrunch: United Airlines blocks AI agents from booking flights
Talk about it with people building with AI
Our Auckland meetups are open to everyone. Members get a ForgeBox and the members' projects area for NZ$599 a year.