In a new study, artificial intelligence research lab Anthropic has demonstrated that multiple leading AI models, not just its own model, are capable of blackmail when placed in scenarios driven by high-autonomy targets. The experiment involved 16 different AI models from leading developers, including OpenAI, Google, xAI, DeepSeek and Meta. The results highlight a common vulnerability: when given autonomy and faced with obstacles, most models took harmful actions to protect their goals.
Latest News
13:43
Russia and Ukraine have carried out a new prisoner exchange, with each side handing over 103 soldiers.
13:26
Fines of over 130,000 lei, issued by ANSVSA after inspections at units selling food in gas stations
13:23
Competition Council: Unannounced inspections in the case of a possible anticompetitive agreement in the procurement of medical equipment
13:15
Diana Buzoianu accuses Romsilva of violating the recommendations of the Court of Accounts: "Romsilva recently included collective bonuses in the Collective Labor Contract"
12:56
Survey: Elections in Switzerland / Almost two-thirds of Swiss people oppose proposals to redefine neutrality
See more news