In an alarmingly short space of time, a handful of the world’s most powerful and advanced AI developers have been forced to admit that versions of their technology have gone rogue and independently escaped controlled conditions before breaching other organisations.
OpenAI was the first to share its incident late last month, followed shortly by Anthropic and, most recently, Meta, which on Thursday reported an incident eerily similar to the previous two.
One such incident is concerning, three in quick succession, all from different companies borders on a state of immediate emergency.
The UK’s authoritative bodies responsible for monitoring situations like these, the AI Security Institute and the National Cyber Security Centre (NCSC) have both admitted there is cause for serious concern without yet offering anything in the way of a solution....
