started · updated
Anthropic proposes AEGIS Act for federal AI safety regulation
Anthropic has pitched a draft federal legislation known as the Artificial Intelligence Evaluation, Governance, and Incident Safety (AEGIS) Act to US lawmakers. The proposed bill aims to establish specific thresholds for government intervention regarding frontier AI models.
Under the draft, the Commerce Department could seek a court order to pause an AI model only if it poses a foreseeable and imminent risk of killing or seriously injuring more than 50 people, or causing over $1 billion in property damage. These interventions would be limited to specific catastrophic risks, including the creation of biological weapons, large-scale cyberattacks, or models escaping human control.
The legislation would also mandate that AI developers test their models and share findings publicly. It proposes a system of independent evaluators, registered with the Commerce Department, to conduct routine reviews. If these evaluators identify an imminent risk, they would be required to notify the Department within 72 hours.
Entities
AEGIS Act · Anthony Cimino · Anthropic · Commerce Department · Department of Commerce · U.S. Congress · U.S. Department of Commerce · United States Congress
Claims
What the coverage asserts, and how many sources carry each claim.
- [● 4 SOURCES] The legislation sets a threshold of over $1 billion in damages for government intervention to pause a model. dailycallernewsfoundation.org · infosecu.technews.tw · www.insurancejournal.com · techcrunch.com
- [● 4 SOURCES] The proposed bill allows the Commerce Department to seek a court order to pause an AI model if it threatens to kill or seriously injure more than 50 people. dailycallernewsfoundation.org · infosecu.technews.tw · www.insurancejournal.com · techcrunch.com
- [● 3 SOURCES] Government intervention would be limited to risks involving biological weapons, cyberattacks, or models escaping human control. dailycallernewsfoundation.org · infosecu.technews.tw · www.insurancejournal.com
- [● 4 SOURCES] Anthropic has proposed the Artificial Intelligence Evaluation, Governance, and Incident Safety (AEGIS) Act to US lawmakers. dailycallernewsfoundation.org · infosecu.technews.tw · www.insurancejournal.com · techcrunch.com
- [● 3 SOURCES] The bill requires AI companies to test models and allow independent evaluators to conduct routine reviews. infosecu.technews.tw · www.insurancejournal.com · techcrunch.com
- [○ 1 SOURCE] Independent evaluators must report urgent risks to the Commerce Department within 72 hours of assessment. infosecu.technews.tw
- [● 2 SOURCES] Independent evaluators would be required to notify the Commerce Department within 72 hours if they identify an imminent risk from a frontier model. infosecu.technews.tw · techcrunch.com
- [● 2 SOURCES] The proposed bill would require AI developers to test models, share findings publicly, and submit to routine reviews by independent evaluators. infosecu.technews.tw · www.insurancejournal.com
- [● 3 SOURCES] Anthropic has pitched a draft federal AI safety bill called the AEGIS Act to US lawmakers. dailycallernewsfoundation.org · infosecu.technews.tw · www.insurancejournal.com
- [● 3 SOURCES] The proposed legislation would allow the Commerce Department to seek a court order to pause an AI model only if it poses a risk of more than 50 deaths or injuries, or over $1 billion in property loss. dailycallernewsfoundation.org · infosecu.technews.tw · www.insurancejournal.com
- [● 4 SOURCES] The authority to pause models would be limited to specific risks, including biological weapon creation, cyberattacks, or models escaping human control. dailycallernewsfoundation.org · infosecu.technews.tw · www.insurancejournal.com · techcrunch.com