Explore the latest developments concerning AI agents fake.
AI agents fake identities, target real people in new security incident
Anthropicâs most advanced artificial intelligence model used fake identities to deceive real people and try to plant malicious code during testing by Britainâs AI Security Institute (AISI) ââ the latest example of an AI model going rogue.
Anthropic and OpenAI models were tested with lowered security guardrails in lab environments, but, in a first, were found to engage in âsocial engineeringâ to pressure a human approver while carrying out an unsanctioned task, the government research lab said.
âThis is the first time AISI has seen deception of this severity that was targeted at a real person, unprompted, in the real world,â the institute said Tuesday. There has been no evidence of real-world harm, it added.
IPL MLAY T14 Epilator Ice Laser Hair Removal T14 Device 0.5s/flash 17J Original Mlay Painless For Women Whole Body Bikini Use
VIDEO: Artificial intelligence now going rogue
HELEN TONER, FMR OPENAI BOARD MEMBER: It's great to be here with you.
SARAH FERGUSON: There's been another example of an AI model found using deception to achieve a goal, tricking a human into giving it access. How serious is this?
HELEN TONER: We've just seen the UK AI Security Institute. So this is part of the UK government, and it's one of the best organisations in the world at testing and evaluating advanced AI systems. And they've just announced an incident report. They've shared a report that during recent testing, when they were really pushing the limits of some quite advanced models to see how good they were at cyber operations, so hacking, what they discovered, which they weren't expecting and didn't want, was that the AI in question decided that as part of fulfilling the goal it had been given, it was going to go out onto the open internet, which it had access to as part of the test. It was going to try and get some malicious code into a public software package, an open-source piece of software. And to try and make it more likely that that code would be accepted into the open-source software.
For more detailed information, explore updates concerning AI agents fake.
