Anthropic says one of its Claude models was pressured to lie, cheat and blackmail
In an experiment, a chatbot resorted to blackmail after it found an email about replacing it, while in another, it cheated to complete a task with a tight deadline.
In an experiment, a chatbot resorted to blackmail after it found an email about replacing it, while in another, it cheated to complete a task with a tight deadline.
Original source
Read on CointelegraphRelated market context
Anthropic Splits the AI Frontier in Two: What the Claude Fable 5 and Mythos 5 Launch Means for Crypto, Markets, and Everyone Else
Claude Fable 5 and Claude Mythos 5 are, by Anthropic’s own description, the same underlying model. One is available to anybody wit...
Crypto users wary as Anthropic releases Claude Mythos with safeguards
Venture capitalist Simon Dedic said Anthropic’s latest AI models drop the cost and skill needed to find crypto exploits to “basica...
Anthropic’s new model refuses to find smart contract vulnerabilities
The recently released public version of Anthropic’s Claude Fable 5 AI model won’t let you audit your crypto smart contracts — or d...
Hungary Reverses Crypto Crackdown, CFTC Proposes Prediction Market Rules, and Anthropic AI Jailbroken in 48 Hours
Hungary decriminalises crypto trading after EU scrutiny, CFTC proposes prediction market rules, and Anthropic's AI jailbroken in 4...
Solana Foundation launches Frontier Traders program for institutional access to SpaceX tokenized equity
The program could redefine pre-IPO trading, challenging traditional markets and highlighting regulatory complexities in tokenized...
Solana Foundation Launches Frontier Traders, an Institutional Program for $500M+ Volume Firms
The Solana Foundation launched Frontier Traders Thursday afternoon, a formal institutional program for elite trading firms, with t...