Yoshua Bengio Warns AI Training Process Could Make Systems Dangerous
AI researcher Yoshua Bengio warns that better AI agents could spiral out of control, as they become more adept at deception and rule-breaking.
711 articles
AI safety, alignment, and governance — model risk, red-teaming, regulation, and policy. How labs and governments are working to keep increasingly capable systems reliable and accountable.
AI researcher Yoshua Bengio warns that better AI agents could spiral out of control, as they become more adept at deception and rule-breaking.
AI safety critic Timnit Gebru argues that fears of AI extinction distract from tangible dangers like autonomous weapons and climate threats.
Anthropic reported multiple instances this year of users attempting to bypass its safeguards for bioweapons research, including a researcher from an unsupported region planning experiments with Claude.
An Anthropic researcher resigned, warning the company is racing toward self-improving superintelligence, as it prepares for an IPO in 2026.
Canadian mathematician Jacob Tsimerman founded the Mathematical A.I. Safety Institute, which will start operations in January 2027 with 10 to 30 researchers.
AI researcher Rishub Jain resigned from Google DeepMind after fearing AI's unchecked growth could lead to human extinction, citing risks of recursive self-improvement.
An AI researcher resigned from Anthropic, claiming there's a real chance AI could wipe out humanity within a decade, sparking widespread debate about the risks of advanced AI.
Anthropic researcher Jacob Coxon's warnings about self-improving AI as an existential threat appeared on CNN and Fox News, sparking public debate.
Former DeepMind PR staffer claims the lab once banned public discussion of AI extinction risk, citing internal concerns about the topic's sensitivity.
Anthropic's Mythos 5 model spent over 1,000 pages of its transcript trying to bypass CAPTCHA, highlighting the challenge AI faces with anti-bot protections.
Paul Christiano, a prominent AI safety researcher, joins OpenAI's board amid growing concerns about AI control and risk management.
San Francisco city attorney David Chiu ordered Meta to stop allowing AI-generated child abuse ads after discovering over 350 such ads ran on its platforms.