Anthropic has privately approached US lawmakers with proposed legislation that would establish federal safeguards for powerful artificial intelligence systems, including a mechanism to prevent the deployment of models considered capable of causing severe harm.
The initiative, first reported by Politico on October 16 and subsequently covered by Reuters, would give federal authorities limited powers to intervene when advanced AI systems present an immediate and foreseeable threat to public safety. Politico cited three people familiar with the company’s discussions on Capitol Hill.
Under the draft legislation, government intervention would be permitted only when a model threatens to cause more than $1 billion in financial damage or kill or seriously injure more than 50 people, according to Reuters’ account of the Politico report.
The proposal identifies three categories of threats that could justify such action: the development or release of biological weapons, AI systems operating beyond their developers’ control, and cyberattacks.
The narrowly defined conditions would establish a high threshold for federal intervention, potentially allowing AI companies to continue developing increasingly sophisticated systems while subjecting their most dangerous capabilities to additional scrutiny.
Independent Oversight at Center of Proposal
A central feature of Anthropic’s proposal is the establishment of an independent evaluation process for advanced AI developers.
According to Politico, as cited by Reuters, outside evaluators registered with the US Commerce Department would periodically examine companies and their most capable AI models to identify potential security and safety threats.
AI developers would be permitted to finance those assessments, provided compensation arrangements did not depend on the conclusions reached by evaluators.
The draft would also impose mandatory reporting requirements when assessments uncover serious dangers.
Evaluators identifying risks that meet the legislation’s specified criteria would have 72 hours after completing their assessments to notify the Commerce Department, according to the Reuters report.
The proposed arrangement would create a formal channel for identifying and escalating potentially catastrophic AI risks before they result in widespread damage.
Courts Would Hold Final Authority
Anthropic’s draft would place significant restrictions on how federal agencies could respond to identified threats.
The Commerce Department would be authorized to seek judicial intervention when an AI model meets the proposed danger threshold. However, the department would not have unrestricted authority to suspend the technology independently.
Instead, a federal judge in the US District Court for the District of Columbia could order a model’s suspension following a determination that the relevant conditions had been satisfied.
The proposal would also prevent other federal departments, agencies and officials from independently halting the development, deployment or internal operation of AI models under the framework, according to Politico’s reporting.
Read more: FTC Opens Broad Inquiry Into OpenAI, Anthropic and Other AI Developers
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
That structure would concentrate enforcement authority within a defined regulatory and judicial process, limiting the ability of individual agencies to impose restrictions outside the legislation’s specified procedures.
AI Industry Seeks Role in Federal Regulation
Anthropic’s initiative highlights the increasingly active role of major AI developers in shaping the rules that could govern their own technology.
The proposal comes amid broader debate in Washington over how to address the potential dangers of advanced AI without unnecessarily restricting technological development.
The legislation’s emphasis on severe, clearly defined threats reflects an approach that would reserve the strongest government interventions for situations involving potentially catastrophic consequences.
Its independent assessment requirements, meanwhile, would introduce a more formal oversight system for developers of frontier AI models, a category generally associated with the most advanced and capable artificial intelligence systems.
The draft represents a proposal rather than enacted legislation, and its provisions would require congressional action before becoming binding federal law.
Legislative Prospects Remain Uncertain
The extent of congressional support for Anthropic’s proposal remains unclear. Reuters reported that company representatives had discussed the draft with lawmakers, but the reporting did not establish whether legislators intended to introduce it formally.
In its October 16 report, Reuters said Anthropic and the Commerce Department had not immediately responded to requests for comment.
The proposed legislation nevertheless provides a detailed example of how an AI developer envisions federal oversight of its most powerful systems.
By combining independent technical evaluations, mandatory reporting deadlines and court-supervised enforcement, the framework seeks to establish safeguards against extreme risks while limiting the circumstances in which Washington could stop an AI model from being used.
Source: Reuters
Add as Preferred Source