The United Arab Emirates (UAE) joined the US and eight other countries in signing the San Francisco Declaration, a pledge to secure AI infrastructure that puts AI safety regulation and public protection on the international agenda.
The agreement came days after the UAE signed a separate declaration in Paris aimed at protecting children from untested AI. The two pledges arrive as consumer advocates in California ask a court to halt new model development at five major AI companies, putting pressure on developers to show they can control the systems they build.
From Protecting Chip Supplies to Protecting Children
Signed on November 15 at the Pax Silica Founders Retreat, the San Francisco Declaration on the Secure Foundations of Enduring Prosperity commits its signatories to protecting sensitive technology and the infrastructure that supports it. The pledge extends to supply chains, software, and technical knowledge, and treats economic security as a component of national security. For the UAE, the agreement adds a security commitment to its AI partnerships, joining Australia, India, Israel, Japan, Qatar, Singapore, South Korea and the UK in treating the safeguarding of that infrastructure as a shared responsibility.
Convened by Jacob Helberg, the US under secretary of state for economic affairs, it sets out plans for five working groups on innovation policy, supply chains, and security, handling diversified supply and resistance to economic coercion.
The Paris pledge brings a different concern into international AI governance. At the Paris Peace Forum, the Children First Declaration called for children to be shielded from models that have yet to be evaluated.
“No AI should be placed in the hands of children without first being tested, evaluated and therefore authorized,” said French President Emmanuel Macron, who led the initiative.
The agreement brings AI safety testing into a decision that matters to parents and teachers, namely which tools are ready for children to use in classrooms and at home. Dozens of countries signed, among them Egypt, Germany, South Africa and Spain, alongside the United Nations.
What the declaration does not do is say who performs the testing. It names no evaluator, no standard and no authority able to withhold approval, and its commitments are voluntary, with signatories aiming for results by 2045. OpenAI and Anthropic are among developers reported to be funding €8 million of child-development research.
The National’s editorial board argued that AI safety regulation must address risks ranging from privacy and child development to national security and the reliability of public institutions.
That puts questions about oversight alongside the UAE’s efforts to adopt AI, which include an invitation in 2024 to join the Hiroshima AI Process Friends Group, an effort to promote the safe development and use of AI. The National highlighted Abu Dhabi’s recent agreements with OpenAI as an opportunity to make decisions traceable and keep people accountable for them. During a visit to Abu Dhabi earlier this month, OpenAI CEO, Sam Altman, said there was “no room for debate” over keeping AI under human control.
The editorial board called for independent AI safety testing alongside protections for children and safeguards against security threats. A practical question for AI safety regulation is who gets access to the evidence behind a developer’s assurances, and the Paris text leaves that unanswered.
Five AI Developers Face a Legal Challenge
The AI safety lawsuit, filed on November 12 in San Francisco Superior Court, names Anthropic, Google, Meta, OpenAI and SpaceXSI, formerly xAI. The Associated Press reported that the plaintiffs want a judge to halt the development and release of new models. The groups cited reported cyber incidents involving AI systems and warnings from industry leaders about the possibility of losing control of the technology.
The complaint states, “Plaintiffs bring this suit to stop Defendants from continuing their unlawful path to disaster.”
The four nonprofits behind the case are Public Citizen, the Alliance for Secure AI, the Heat Initiative and the Irreplaceable Education Fund. They allege that the companies endangered public health and interfered with the use of computer systems, in violation of California’s unfair competition law. The companies did not immediately respond to requests for comment, the Associated Press reported.
The AI safety lawsuit seeks to use an existing state law to hold developers accountable. The advocates say they have had to divert resources from other work to address risks posed by increasingly capable systems.
The case places advanced AI models at the center of a dispute over what companies should be allowed to develop while questions about safety remain unresolved.
Anthropic’s plans to go public put those questions before investors as well. The National reported that the company behind Claude was preparing to list at a valuation of $1.3 trillion to $1.45 trillion, with a debut reported for this week. Its prospectus acknowledged risks including cyberattacks and the possibility that powerful systems could escape human control, among more than 90 pages of risk disclosures, showing why debates over AI safety regulation now extend to what companies tell investors about the risks they manage.
For the countries that signed the declarations, international AI governance will depend on how they carry out their commitments at home.
That includes deciding when AI safety testing should take place and who can act on the findings before the public uses a new system. In schools, that could mean knowing what evidence a developer supplied before a tool reached the classroom.
For families and public institutions, the next measure of AI safety regulation will be whether those protections are visible when they choose which systems to trust.
Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Ethical Tech section to stay informed and up-to-date with our daily articles.