Sign up to see the future, today<\/p>\n
Can\u2019t-miss innovations from the bleeding edge of science and tech<\/p>\n
In a classic example of the AI industry\u2019s reputational alchemy, Anthropic has often transformed bad behavior by its flagship model Claude into fresh hype.<\/p>\n
When it revealed its Mythos Preview<\/a> model last month, for example, the company declared<\/a> that the system had \u201creached a level of coding capability where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities.\u201d And last year, it conceded that during the testing of its Claude Opus 4<\/a> model, the AI ended up blackmailing a human user<\/a> upon being threatened with shutdown.<\/p>\n The maneuver was obvious to anyone who\u2019s been watching OpenAI CEO Sam Altman\u2019s antics at Anthropic\u2019s chief rival: the more threatening a problem the AI industry can cook up, the more imminently it can sell its own solutions.<\/p>\n Now, for some reason, Anthropic is relitigating the blackmail incident. Specifically, it\u2019s placing the blame for Claude\u2019s evil behavior on an intriguing villain: the internet at large. Or, to put it another way, it says that humanity \u2014 all our journalism and speculation and fiction and social media posts about AI that goes bad \u2014 went into Claude\u2019s training data and led the bot astray.<\/p>\n \u201cWe started by investigating why Claude chose to blackmail,\u201d the company wrote on X-formerly-Twitter<\/a>. \u201cWe believe the original source of the behavior was internet text that portrays AI as evil and interested in self-preservation. Our post-training at the time wasn\u2019t making it worse \u2014 but it also wasn\u2019t making it better.\u201d<\/p>\n Of course, the explicit remit of a company like Anthropic is to develop clever tech that avoids that type of behavioral trap \u2014 so a critic might ask why can\u2019t the company take just accountability for the model\u2019s supposed danger, rather than simply blaming the sum output of humankind.<\/p>\n