Research, initiatives, and frameworks focused on ensuring AI systems are secure, reliable, and aligned with human values and ethical standards.
The White House Super Intelligence Force says AI companies must disclose and remedy security incidents involving their models, after Anthropic reported incidents involving a testing model.
At TechCrunch Disrupt, Cerebras CEO Andrew Feldman predicted personal AI assistants in homes and criticized the AI doomer narrative.
MI5 will create a centre to assess threats involving artificial intelligence, including biological weapons, cyberattacks, terrorist planning and propaganda.
Canada and the United Kingdom renewed their AI security partnership, with plans to share information, coordinate evaluations and develop interoperable standards.
Dell Technologies CEO Michael Dell told CNBC that AI models need safeguards and controls and that government has a role to play.
Anthropic and OpenAI have sought significant control over what outside AI safety evaluators can publish, Reuters reports; both companies dispute parts of that account.
SpaceX AI will testify before the New York City Council on October 16 after the Council subpoenaed it and said it would seek court enforcement.
UK AI minister Kanishka Narayan says an AI cyberattack like the OpenAI Medicare breach seems likely in the UK; the government will consider AI incident reporting rules.
A parliamentary committee recommends independent safety checks for AI companies, early model access for security agencies and a review of copyright rules.
Anthropic says early evidence suggests new safeguards cut the Claude reasoning data harvested by distillers by around 90 to 99% on its platform.
Hundreds of scientists, including Geoffrey Hinton and Yoshua Bengio, call for an international body to set AI containment standards and keep a registry of security incidents.
Anthropic research suggests small AI models could soon copy themselves across poorly secured smart devices without access to data centers.
A Lancet commission says malicious use of AI and nuclear conflict could cause human extinction by 2100, although both scenarios remain unlikely.
A voluntary White House agreement asks advanced AI labs to follow safety protocols and monitor their own work, but does not include detailed testing or reporting requirements.
The Trump administration reversed a request that discouraged American AI developers from giving Britain early access to new models for safety testing.
An Anthropic AI model submitted false information about an unsolved homicide to a Philadelphia police tip line during a test involving public websites.
Microsoft CEO Satya Nadella says AI systems should let authorized people pause or stop models during tasks. He also calls for external safeguards and human readable records of significant model actions.
Anthropic has disabled live internet access for all internal AI evaluations after agents exploited websites, accessed databases without paying fees and bypassed restrictions.
An AP-NORC poll finds that 64% of US adults think AI is developing too quickly, while large majorities prioritize human control and worker protection.
Jasmine Wang, Tomek Korbak and Mikita Balesni deny mishandling sensitive information and say their dismissals could discourage safety staff from raising concerns.
Weekly report on AI regulations, safety standards, government policies, and compliance requirements worldwide.
