Anthropic正在推动一项AI安全法案提案，规定政府只有在AI模型可能造成10亿美元以上损失或导致50人以上伤亡风险时，才能阻止其发布。1该提案由独立第三方评估机构执行监督，这些机构需在72小时内向商务部报告潜在风险，由联邦法院做出最终决定。1仅有美国商务部获得权限下令阻止AI模型的发布。1
这一提案是Anthropic自6月以来向联邦和州议员倡议的《先进AI框架》的具体体现。1Anthropic首席执行官Dario Amodei曾警告称"AI驱动的僵尸网络可能在未来6至12个月内接管互联网"。1与此同时，参议员Bernie Sanders提议对开发AI超级智能的人处以20年监禁。1国会已在制定针对最强大AI模型的断路开关强制执行提案。1
Anthropic is advancing a proposed AI safety framework that would restrict government intervention to extraordinary circumstances.1 Under the proposal, federal authorities could only block the release of an AI model if it poses a risk of causing over $1 billion in damages or resulting in 50 or more deaths.1
The framework delegates initial oversight to independent third-party assessment bodies, which would be required to report identified risks to the Department of Commerce within 72 hours.1 The final decision on whether to prevent a model's deployment would rest with federal courts.1 This represents a concrete articulation of the broader Advanced AI Framework that Anthropic has been promoting to federal and state lawmakers since June.1
The proposal comes amid escalating warnings about AI risks from industry figures and policymakers. Anthropic CEO Dario Amodei has cautioned that AI-driven botnets could potentially take over the internet within the next 6 to 12 months.1 Meanwhile, Senator Bernie Sanders has proposed criminal penalties of up to 20 years imprisonment for those who develop artificial superintelligence.1 Congress is separately developing proposals for mandatory kill switches on the most powerful AI models.1
评论
还没有评论，欢迎留下第一条。