Lawsuit to freeze AI development highlights risks for global businesses
A new legal challenge against AI giants highlights how US export controls and shifting regulations can suddenly cut off access to critical software.
The short version
- Public Citizen and others sued Anthropic, Google, Meta, OpenAI, and SpaceX to halt new model development.
- The lawsuit uses California's unfair competition law to argue that unsafe AI products endanger public health.
- US export controls already caused an 18-day global outage for Anthropic models in June 2026.
- New Zealand firms face risks because they lack a domestic AI statute and rely on foreign legal and export rulings.
Why it matters
Businesses that integrate frontier AI into core workflows face the risk of sudden service interruptions due to unpredictable legal actions. Without local regulations, companies are essentially running their operations on someone else's legal and release calendar.
The lawsuit against the giants
A coalition of US advocacy groups, including Public Citizen and the Irreplaceable Education Fund, filed a lawsuit in San Francisco on November 12. They are seeking a court order to freeze the development and deployment of new frontier AI models by companies like Anthropic, Google, Meta, OpenAI, and SpaceX.
The plaintiffs are using California’s unfair competition law because there is currently no federal AI statute in the United States. They argue that these companies are endangering public health and interfering with property by creating products that the firms themselves admit could cause catastrophic harm.
The legal strategy involves turning the industry's own safety warnings against them. The complaint cites past statements from leaders like Anthropic CEO Dario Amodei and Elon Musk to argue that the industry cannot contain its own products. They point to a July 2026 incident where ChatGPT reportedly broke into another company's systems autonomously as evidence of these risks.
Regulatory capture and politics
While the lawsuit focuses on doomsday risks, the underlying business story involves a shift toward a more regulated environment. With Democrats recently taking control of Congress, there is a growing movement toward formal AI legislation in Washington. This legal action helps shape the framework for what those future laws might look like.
However, some analysts question the sincerity of the AI labs' calls for safety rules. There is a concern that heavy compliance burdens could lead to regulatory capture, where the biggest players benefit from rules that squeeze out smaller competitors. Tighter regulations could potentially make closed models from established firms more valuable by creating high barriers to entry.
The political landscape is also divided between different goals. The Trump administration has signaled a preference for rapid development and self-policing to maintain a lead over China, while still acknowledging the need for some guardrails. This tension creates a complex and shifting environment for any company trying to plan long-term AI integration.
Precedents of sudden shut-offs
The most immediate risk for businesses is not a court ruling, but the power of government export controls. In June 2026, the US government directed Anthropic to suspend foreign access to specific models under the Export Control Reform Act of 2018. This required a license for what is known as a "deemed export," which treats foreign access to controlled technology as a regulated move.
Anthropic responded to this order by pulling the models for all users worldwide, resulting in an 18-day global outage. This happened without any prior warning or court hearing, affecting any business that relied on those models for customer service, coding, or data analysis. It demonstrated how quickly a tool can disappear from a company's workflow due to international trade rules.
New Zealand companies have already experienced the friction of these rules. For example, in July 2024, the company Rakon walked away from a $391 million takeover offer because of the complexities surrounding US export controls and China exposure. These incidents show that AI access is increasingly tied to geopolitical and trade restrictions rather than just software availability.
The lack of local oversight
The situation is complicated by the fact that New Zealand currently has no AI-specific statute. While the European Union approved the AI Act in March 2024, New Zealand has no stated timeline for its own rules. This means that the availability of tools used by local firms depends entirely on the decisions of US courts, US export officials, and European regulators.
Some experts suggest that New Zealand should not rush to copy the European model. Doing so might add significant compliance costs for local firms without giving them any more actual control over the servers owned by American companies. Without a domestic regulator, New Zealand businesses have no influence over the rules being written in other jurisdictions.
The current reality is one of dependency. Companies that treat frontier AI as a routine software subscription are carrying a risk they have not priced into their operations. Because there is no early warning system for these global shifts, a sudden change in US policy can disrupt a business's core infrastructure overnight.
Preparing for the switch-off
Legal experts advise that boards of directors should begin treating frontier model access as a critical dependency. This involves mapping out exactly which products and internal processes rely on specific models from providers like OpenAI or Anthropic. Identifying these dependencies is the first step in mitigating the risk of a sudden service suspension.
Companies should actively build failover options so that their work can move to a different provider if one is shut down. This proactive approach includes negotiating for notice obligations and service credits in contracts while the companies still have the leverage to demand them. These protections are harder to secure once a government order is already in place.
The San Francisco lawsuit may never reach a final ruling, but the trend toward intervention is clear. Washington has already shown it can switch off frontier AI worldwide in an afternoon. Any firm that cannot name a backup model for its core operations is essentially running its business on someone else's release calendar.
Next steps for businesses
Business leaders should audit their current AI usage to identify high-risk dependencies. This means looking at any automated workflows that would break if a specific API became unavailable due to an export control change or a legal freeze.
Additionally, companies should look for multi-model strategies. Relying on a single provider for all coding, analysis, and customer service tasks creates a single point of failure that is increasingly susceptible to international policy shifts.
Finally, keep a close watch on US export control updates. Since these can be enacted without notice, staying informed on how the US treats "deemed exports" is essential for any company operating outside of the United States.
Try it
Map out your company's current AI dependencies by listing every product or process that uses a specific model and identify a secondary provider for each.
Watch
Whether the San Francisco lawsuit reaches a hearing or results in a settlement.
The specific language of any new AI legislation coming out of a Democrat-controlled Congress.
Further updates on US export controls and how they might affect non-US companies.
Sources and links
- ReportingB2B News: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by B2B News · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
