started · updated
Anthropic proposes US federal bill to regulate frontier AI safety
Anthropic has proposed a federal legislative framework to the United States to regulate frontier artificial intelligence safety and security. The draft bill seeks to establish high thresholds for government intervention, suggesting that the government should only be permitted to block an AI model if it poses an imminent risk of causing more than 50 deaths or serious injuries, or economic damages exceeding $1 billion.
Under this proposal, the Department of Commerce would hold primary authority over suspending AI development or deployment, with any such action requiring authorization from a federal judge in the District of Columbia. To ensure oversight, the plan requires frontier AI developers to undergo periodic evaluations by independent, third-party evaluators registered with the Department of Commerce.
This legislative push comes amid growing scrutiny of AI safety. The Federal Trade Commission has issued civil investigative demands to both OpenAI and Anthropic regarding security incidents and safeguard practices. Additionally, researchers have expressed significant concerns regarding existential risks, with some surveys indicating that a segment of the scientific community estimates an 18 percent probability of human extinction resulting from future AI advancements.
Entities
Anthropic · Capitol Hill · Claude · Dario Amodei · Department of Commerce · District Court for the District of Columbia · Federal Trade Commission · OpenAI · Politico · U.S. District Court for the District of Columbia · United States · United States Congress
Claims
What the coverage asserts, and how many sources carry each claim.
- [● 7 SOURCES] The proposal requires frontier AI developers to undergo periodic evaluations by independent evaluators registered with the Department of Commerce. columnadigital.com · srnnews.com · blockonomi.com · china.timesofnews.com · www.newsbytesapp.com · +2 more
- [● 6 SOURCES] The bill limits intervention to risks involving biological weapons, cyberattacks, or models evading developer controls. srnnews.com · blockonomi.com · china.timesofnews.com · www.cryptopolitan.com · www.newsbytesapp.com · +1 more
- [● 9 SOURCES] The proposed bill allows the government to block an AI model only if it poses an imminent risk of more than $1 billion in damages or more than 50 deaths or serious injuries. columnadigital.com · srnnews.com · arynews.tv · blockonomi.com · china.timesofnews.com · +4 more
- [● 5 SOURCES] The bill would prohibit any US government department or agency other than the Department of Commerce from suspending AI development or deployment. srnnews.com · blockonomi.com · china.timesofnews.com · www.world-today-news.com · www.it-boltwise.de
- [● 5 SOURCES] Under the draft, only a federal judge for the District Court for the District of Columbia could authorize the suspension of an AI model. srnnews.com · blockonomi.com · china.timesofnews.com · columnadigital.com · www.cryptopolitan.com
- [● 11 SOURCES] Anthropic has pitched a draft federal bill to US lawmakers to regulate frontier AI safety and security. columnadigital.com · srnnews.com · arynews.tv · blockonomi.com · china.timesofnews.com · +5 more
- [● 2 SOURCES] The proposed law would trigger a block if a model is expected to cause more than 50 deaths or injuries, or damages of at least 1 billion US dollars. www.it-boltwise.de · columnadigital.com
- [○ 1 SOURCE] Anthropic reported blocking targeted misuse attempts by state and politically motivated actors. www.berliner-zeitung.de
- [● 2 SOURCES] Anthropic is advocating for an AI safety law that would only permit the blocking of a model in extreme damage scenarios. www.it-boltwise.de · columnadigital.com
- [○ 1 SOURCE] Anthropic identified instances where users attempted to use the Claude model to modify the Chikungunya virus or adapt bird flu to mammals. www.berliner-zeitung.de
- [○ 1 SOURCE] Meta removed a clause from its commercial terms of use that would have allowed the systematic scraping of news content for AI training. borncity.com
- [○ 1 SOURCE] The use of unauthorized AI tools increased the average cost of a data breach incident by 670,000 US dollars. www.ad-hoc-news.de