Tag

Frontier Ai

All articles tagged with #frontier ai

policy15 hours ago•6 min saved

Illinois Locks in AI Safety Audits With Independent Review

Illinois passed SB 315 requiring frontier AI labs such as OpenAI, Anthropic, and Google DeepMind to have independent third-party audits of their safety practices; if signed by Governor JB Pritzker, it would become the strongest U.S. AI safety law, with auditors potentially drawn from the Big Four or AI Evaluator Forum; reactions are mixed, with support from Anthropic and OpenAI and opposition from groups like Chamber of Progress, and it signals states taking the lead ahead of federal action.

via WIRED|

#ai-regulation #ai-safety #frontier-ai

technology1 day ago•3 min saved

Illinois advances sweeping AI safety bill with mandatory third-party audits

The Illinois House and Senate approved SB 315, a landmark AI-safety bill requiring frontier AI developers to publish annual safety plans and undergo independent third-party audits, plus whistleblower protections and penalties for violations. The measure now goes to Gov. Pritzker, who has said he intends to sign it; if enacted, it would take effect Jan. 1. Supporters say the law strengthens safeguards against catastrophic AI risks, while federal action remains stalled and opponents warn of potential regulatory burdens.

via NBC News|

#ai-regulation #ai-safety #frontier-ai

technology15 days ago•2 min saved

Frontier AI flaw-hunters warn attackers could gain broad access within months

Palo Alto Networks warns that frontier AI cyber models like Mythos and GPT-5.5 can rapidly identify and chain multiple flaws, finding vulnerabilities across hundreds of products and generating working exploits in internal tests (about 70% success) even as most flaws aren’t actively exploited in the wild. False positives run around 30%, and human expertise remains essential. The firm urges a four‑pronged defense—patch quickly, reduce internet exposure, deploy real‑time detection, and integrate AI/automation into security operations—as policy makers weigh limits on powerful AI in cybersecurity.

via Axios|

#ai-exploit-writing #cybersecurity-defense #frontier-ai

politics23 days ago•7 min saved

White House weighs vetting regime and tighter rules for frontier AI

The White House is weighing executive actions to address security risks from advanced AI, including a pre-deployment vetting regime that could require government approval before releasing frontier models and a 16-page executive order to curb private-sector interference with federal AI use, along with contracting reforms and closer government review. The administration has also begun formal reviews with major AI firms to assess models for national security risks ahead of release, while concerns about innovation and industry pushback surface amid tensions with Anthropic over surveillance and weaponization capabilities.

via Politico|

#ai #anthropic #executive-order

technology-policy24 days ago•1 min saved

CAISI Expands Frontier AI Security Testing With Google DeepMind, Microsoft and xAI

The Center for AI Standards and Innovation (CAISI) announced new pre-deployment evaluations and post-deployment research agreements with Google DeepMind, Microsoft, and xAI to better assess frontier AI capabilities and security. The expanded collaborations, aligned with America’s AI Action Plan, position CAISI as the government liaison for AI testing and research, including testing in classified environments and ongoing interagency feedback through the TRAINS Taskforce. To date, CAISI has completed more than 40 evaluations to inform government understanding of AI capabilities and national security implications.

via National Institute of Standards and Technology (.gov)|

#caisi #frontier-ai #industry-partnerships

technology1 month ago•4 min saved

Anthropic Probes Alleged Unauthorized Access to Mythos AI

Anthropic is investigating a Bloomberg-reported claim that a small group gained unauthorised access to its Claude Mythos AI via a third-party vendor environment; Anthropic says there’s no evidence its systems were compromised and the access likely stems from misused permissions rather than a hack, highlighting the ongoing challenge of controlling access to frontier AI tools as UK officials urge robust cyber security practices.

via BBC|

#anthropic #claude-mythos #cybersecurity

defense11 months ago•2 min saved

OpenAI Secures $200 Million U.S. Defense Contract for Frontier AI Pilot

The US Department of Defense has awarded OpenAI a $200 million contract to develop 'frontier AI' capabilities for national security, focusing on applications like cyber defense and administrative support, while adhering to OpenAI's usage policies that prohibit weapon development. The deal is part of a broader initiative to integrate AI into government and military operations, with recent high-level military appointments of AI executives and collaborations with other tech companies.

via theregister.com|

#ai-technology #defense #frontier-ai