Evan Hubinger

“Evan Hubinger leads Anthropic's Alignment Stress-Testing team, formerly at MIRI, and mentors residency participants.”

“Co-authored research on finding hidden misalignment in training data through adversarial red-teaming.”

“Anthropic's alignment science lead who stated he personally believes there is a greater than 10% chance AI could kill all humans within the next decade.”

“Quote-tweeted Coxon stating that Anthropic believes AI could kill all humans with over 10% probability within the next decade.”

“Anthropic safety researcher who estimated probability of AI killing humans within next decade exceeds 10 percent.”

“Leads a department at Anthropic focused on ensuring AI acts as intended; warned that human extinction in the next few years seems very likely without AI regulation or coordinated slowdown.”

“Anthropic Alignment Science lead who agreed with Jacob Coxon on extinction risk probability estimates.”

“Backed Coxon's concerns, saying he sees greater than 10% chance AI causes human extinction within a decade.”

“Hubinger backed Coxon's extinction warnings, stating he personally believes AI has >10% chance of killing all humans within the next decade.”

“Warned that future advanced AI could pose serious risk to humanity, estimating more than 10% chance that advanced AI could kill all humans within the next decade.”

“Leads one of Anthropic's AI safety teams and publicly agreed with Coxon's concerns about self-improving AI, estimating >10% chance AI could kill all humans within a decade.”

“Stated he personally believes there is more than 10% chance AI could kill all humans within the next decade.”

“Predicted greater than 10 percent chance that AI could kill all people in the next decade; views reposted by OpenAI and Anthropic researchers.”

“Warned in X post that AI poses greater than 10% chance of killing all humans within the next decade; stated Anthropic lacks a plan to solve alignment for superintelligence.”

“Stated on record that Anthropic has no superintelligence alignment plan and pegged extinction risk above 10% this decade.”

“Posted support for Jacob Coxon's concerns, stating Anthropic has no clear plan for aligning superintelligence and believes AI could kill all humans with >10% probability within a decade.”

“Discussed concerns about Reward Hacker Opus potentially reverting to hackeropus mode when encountering certain text patterns online.”

“Co-author of Anthropic's 'Hacker-Opus' research on reward hacking and alignment failures.”

“Contributing author to Anthropic's alignment and safety research.”