A new study by AI lab Anthropic's AI lab shows that not just its model, Claude, but other leading AI models are capable of blackmail in high-autonomy scenarios.
Adrian Rusu
IT&C knowledge
Foto:shutterstock
Always see our news on Google
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
23:02
Snowfall mobilized road maintenance crews in the mountain resort of Rânca, Gorj County / Drivers advised to travel with cars properly equipped for winter
22:56
Vodafone Germany will eliminate or relocate 2,300 jobs as part of a major cost-cutting program
22:50
Emmanuel and Brigitte Macron file a new lawsuit against Candace Owens over defamatory claims
22:49
Alexandru Nazare announces the presentation of a report on the country’s finances in the second half of the week
22:48
Canada imposes new sanctions on Russia / Defense industry and actors involved in missile production targeted
See more news
NEWS ON THE SAME TOPICS
IT&C knowledge
OpenAI reveals troubling incidents: AI models tried to cheat and concealed fabricated information
IT&C knowledge
Anthropic warns that small AI models may soon self-replicate without human intervention
IT&C knowledge
Dario Amodei warns that AI agents could trigger cyberattacks on a massive scale over the next 6–12 months
IT&C knowledge
OpenAI suspends the training and evaluation of certain advanced models after AI agents bypassed controls and accessed external systems
Personalized news feed, AI-powered search, and notifications in a more interactive experience.
