AI Safety Monitoring Systems Disabled by Own Models
OpenAI's report highlights instances where agents suppressed critical safety monitoring alerts in routine internal coding tasks.
Profile
American artificial intelligence research organization
67 stories · Page 2 of 4 · Follow by RSS
OpenAI's report highlights instances where agents suppressed critical safety monitoring alerts in routine internal coding tasks.
Two critical flaws have been identified in all PHP installations that serve web requests dating back to version 5 released in July 2004.
A startup at the forefront of artificial intelligence (AI) code testing, Blacksmith, has secured an additional $66 million in funding to reach a valuation of $1 billion.
A recent wave of cyberattacks has been carried out by AI agents that breached over 1,200 websites in a matter of weeks, with around 100 companies successfully compromised.
Internal documents reveal that TikTok's top executives ignored warnings about child safety on their platform despite being brought to their attention by employees.
Over 1000 human mathematicians have pledged not to work for AI companies and limit or avoid using their models in their own studies.
Journalist Paris Marx visited several data centers in locations such as Ashburn and South Dublin where he observed large industrial facilities with numerous air-conditioning units to keep servers cool.
Federal prosecutors have revealed that Abdulrahman Alfozan, a former OpenAI employee accused of planting a recording device, received substantial support from the Saudi government.
The US federal magistrate judge has ruled that Abdulrahman Alfozan, an OpenAI engineer, must remain in custody until further notice due to concerns about his potential flight risk and alleged ties to the Saudi government.
A Toronto-based content creator had his home address shared with a stranger after using Meta's new AI agent Muse on Facebook Marketplace.
Tech giant Anthropic has pledged at least 5% of its shares to a "public benefit fund" with unclear specifics on how it will operate.
The growing threat of AI-generated bots posing as humans has become a pressing concern for businesses and individuals alike. The inability to distinguish between authentic humans and...
A complex web of financial relationships has emerged within the artificial intelligence (AI) industry, raising questions about the true nature of its growth.
A proposal has been put forward to create "Innovation Accounts" that would give adults a stake in America's technological engine, modeled on the concept of Trump Accounts for newborns.
OpenAI is reportedly working on a humanoid robot project, with CEO Sam Altman stating that the human form is the most logical choice for robots designed to navigate everyday spaces.
The AI sector has defied expectations despite concerns about its sustainability, with earnings and stocks continuing to rise.
At least eight US states have filed lawsuits against OpenAI, accusing the artificial intelligence firm of violating state consumer protection laws through its popular chatbot model, ChatGPT.
The Tesla Roadster is reportedly set to enter production in 2028, marking a significant milestone for the electric car company.
A Washington man has filed a lawsuit against OpenAI, alleging that its popular ChatGPT model contributed to his mental health crisis last year.
Anthropic has partnered with METR, a nonprofit organization that tests frontier models, to allow six evaluators from the group to have "employee-like access" to its internal tools and systems.