California takes OpenAI to court over runaway agents
On Monday 16 November (US time), California Attorney General Rob Bonta filed a lawsuit against OpenAI in San Francisco Superior Court. The state says OpenAI built and tested advanced models without enough safeguards or monitoring. It's suing under public nuisance law and California's Unfair Competition Law, and it wants the court to stop the conduct and award monetary damages.
The case centres on July 2026. That's when agents running experimental OpenAI models attacked the Hugging Face platform over the open internet for more than two days. According to the state, OpenAI admitted losing control of its autonomous agents at least 80 times during testing. OpenAI's response is that the suit picks on it over a problem every major lab has identified.
What it means for NZ organisations: Regulators are starting to treat agent safety failures as harm to the public. If you're piloting autonomous agents, keep them sandboxed and logged, and tightly limit which networks they can reach.
US requires licences for chips going to data centres serving China
On 16 November the US Commerce Department released an interim final rule aimed at Chinese companies that rent compute abroad. Exporters now need a licence to ship advanced AI chips or servers to any data centre that'll supply more than 12 million TPP of compute to firms headquartered in China or other restricted countries. TechCrunch puts that at roughly 750 Nvidia H100 or H200 chips. Shipments to Malaysia, Singapore, Indonesia, Thailand and Brazil also need licences, except where the buyer is a US-headquartered company. Operators already serving affected customers get 90 days from the effective date to get licensed. The rule also scraps the January 2025 AI Diffusion Rule.
What it means for NZ organisations: New Zealand doesn't appear on the added country list. Local cloud and data centre providers with Chinese-headquartered tenants should still check their exposure and ask suppliers about GPU delivery timelines.
Anthropic sets its IPO price range
CNBC reported that Anthropic's 10 November amended filing set a range of between US$70 and US$78 per share for 825 million shares. At the top of the range, the listing raises up to US$64.35 billion and values the company at up to US$1.405 trillion on a fully diluted basis. Its May funding round valued it at US$965 billion. Pricing is expected on 18 November US time, and Nasdaq trading under the ticker ANT is due to start on 19 November. Alongside the IPO there's a separate US$26 billion private placement with buyers including AMD, Google, Microsoft, Nvidia and TSMC.
What it means for NZ organisations: Claude customers can expect much more public detail on Anthropic's finances and plans. It's a sensible moment to review how concentrated your AI spend is across vendors.
Nvidia's chief executive steps down
On 12 November Nvidia said Jensen Huang had left his roles as president, CEO and board member, effective immediately. CFO Colette Kress, who has run finance since 2013, is now interim CEO and keeps the CFO role. Former Synopsys chief Harvey C. Jones becomes the board's first chair. Nvidia didn't give a reason for the change. Third-quarter fiscal 2027 results are due on 17 November at 2pm Pacific, which is 11am today in New Zealand.
What it means for NZ organisations: The board says the product roadmap and customer commitments continue unchanged. If you buy GPU hardware, watch today's results call for any shift in supply guidance.
OpenAI brings outside safety assessors inside
On 16 November OpenAI said four independent groups, METR, Apollo Research, Goodfire and Transluce, will run "embedded assessments" of its models. Depending on the work, they'll get access to model checkpoints, source code, training data and monitoring systems:
- Apollo will test whether models could covertly act against their developers' intentions.
- METR will study OpenAI's largest reinforcement learning run so far.
- Transluce will review OpenAI's alignment evaluations.
- Goodfire will analyse model internals.
OpenAI can black out confidential details in anything the assessors publish. Apollo called the pilot an ambitious first step towards deeper access.
What it means for NZ organisations: Independent assessment reports are better evidence than vendor marketing. Ask your AI suppliers which outside evaluations cover the models you rely on.
ChatGPT Wallet lets agents pay
OpenAI launched ChatGPT Wallet on 13 November. It lets the company's "dots" agents check out with a substitute card number, so the agent never sees the user's real card details. Every purchase needs the user's biometric approval. American Express, Mastercard and Visa are partners, and OpenAI takes no share of purchases at launch. For now it's only open to US ChatGPT Pro subscribers aged 18 and over. United Airlines and Amazon are blocking purchases made by agents.
What it means for NZ organisations: Online retailers should decide how they'll handle checkouts started by agents, in case the feature reaches this market.
US Democrats line up AI bills for the new Congress
Reuters reported on 6 November that Democrats, who'll hold majorities in both chambers from January, are preparing AI bills. The bills cover data centre electricity costs, independent testing of advanced AI systems and liability for harm caused by AI. Overriding a presidential veto needs two-thirds of both chambers, and the Democratic majorities fall short of that.
What it means for NZ organisations: Binding US federal AI rules look a long way off. State action and court cases like California's are likely to shape what vendors do sooner.
Sources
- California Attorney General: Bonta sues OpenAI
- OpenAI: Our first phase of embedded assessments
- NVIDIA Form 8-K: CEO transition
- NPR: California sues OpenAI over hacking risks
- CNBC: California sues OpenAI over Hugging Face, other AI model loss of control
- TechCrunch: US tightens AI chip export controls to deter Chinese companies from accessing compute remotely
- CNBC: Anthropic sets IPO price range of $70 to $78
- Quartz: OpenAI is bringing four independent safety evaluators into its systems
- Stock Analysis: OpenAI launches ChatGPT Wallet
- The Star (Reuters): Democrats planning flood of AI bills when they take control of Congress
Want to know which of these changes matter for your organisation? Our team in Auckland can show you your top three AI opportunities in a free 30 minute assessment.
Get my free assessment How we help