Amid increasing friction between AI companies and the U.S. government over regulation/export control of frontier AI models, tech giant Google has proposed establishing a frontier AI regulatory organization (FARO) to establish national and international AI safety standards.
In its whitepaper titled “A pragmatic approach to AI governance in America” by Kent Walker, President of Google’s Global Affairs, a framework has been suggested which operates under federal oversight and establishes guardrails to protect the public from the perils of frontier AI models.
Google pointed out in its paper that the country has solved similar challenges in the past through independent, industry-funded bodies which operate under federal oversight and establish guardrails to protect the public. According to the company, FARO would set guidelines for developers to identify, reduce risks and ensure that companies implement security practices and incident response plans before releasing frontier models publicly.
As per Google, FARO can oversee an independent procedural audit regime for frontier AI companies. Such audits can be managed by well-known professional audit organizations on an annual basis and the reports should be provided to the FARO. The company highlighted that as far as AI applications like chatbots are concerned, they raise social and consumer safety issues which are different from the national security issues raised by the most advanced frontier AI models.
It stated that for these widespread applications the government can frame and in some cases amend existing laws and rules to address consequences and specific harms.
The whitepaper by Google comes in the wake of U.S. government putting export control over Anthropic’s Claude Fable and Claude Mythos models and OpenAI GPT 5.6, citing national security concerns over the advanced capabilities of these AI models.
Weeks later, the U.S. government partially lifted the ban on Claude Fable and Claude Mythos and GPT 5.6 allowing its use for selected American companies and employees of the two companies.
The paper recommended that frontier AI safety and security policy should include three key elements. The first should be an independent regulatory organization that can keep pace with fast-moving AI research and development. The second should be defined scientific benchmarks for identifying frontier capabilities in the cyber and chemical, biological, radiological, and nuclear (CBRN) domains, complemented by clear safety and security standards for building, testing, and deploying the most advanced AI systems.
The third Google proposed that frontier AI models should undergo annual audits to ensure they comply with AI safety standards. The audits would verify whether companies are following safety and security procedures which will make these models more transparent.
It flagged uncertainty regarding the impact of AI on jobs and the economy but it suggested that with right policy frameworks and investments in American workers, AI can enhance the capabilities of the workers, which will lead to economic growth and social welfare.
Google is also working closely with partners to equip the workforce, including a manufacturing vocational partnership, and initiatives to train over 300,000 American workers and all six million US educators. It has encouraged policymakers to support private-sector, “earn-and-learn” training models.
“Funding employer-driven upskilling can offset the financial risk for businesses while helping workers acquire the specific, transferable skills that industry demands,” the company mentioned in its paper.
“Policy responses can encourage companies to invest in apprenticeship programs and skilling employees even when those employees may use those skills to find roles with other firms,” it added. The company further informed that it has endorsed a range of bills at the federal level that would help create these opportunities and incentives.
The second challenge as per the paper towards which the government should look upon is protecting kids and families in the digital world.
“The widespread adoption of generative AI chatbots in the US has highlighted concerns around potential harms to our children. AI developers and deployers should offer high-quality, privacy-protective, and age-appropriate AI experiences that empower kids, protect them from inappropriate content, and recognize their developmental needs,” the company argued in its paper adding that the rules should empower parents to assess and address the needs of their children.
It also advised that federal regulations require chatbots to integrate robust suicide and self-harm protocols directly into AI systems. For instance when a user searches or prompts for concerning topics the chatbot should pause the query and present “crisis resource pages,” coordinating directly with lifeline services.
Also Read: Five Eyes Intelligence Warns Against Frontier AI Even as U.S. Gatekeeps Claude Mythos and Fable
