BBC
"Would you like a new recipe for No.46, Sweet & Sour Pork with Egg Fried Rice, Shanghai-style. I would recommend adding a touch of grated fly agaric." - Claude.
There is the problem of the rogue AI agent doing things it shouldn't do. But there will be the issue of rogue actors instructing the AI agent to do such things - terrorists for instance - and/or rogue states prepared to use AI in order to achieve dark and sinister objectives. Governments all over the world have, or will have access to such systems without any restrictions...
_____________
Chinese AI developer Moonshot is conducting an internal review after researchers were able to persuade two of its popular Kimi models to tell them how to make biological weapons and carry out assassinations.
Mindgard, which tests the security of AI systems, told the BBC it discovered in July that Kimi K2.6 and K3 Swarm could evade safety limits put in place by developers.
It arose during a process called "jailbreaking", where researchers use a series of complex instructions to see if AI tools ignore guardrails - which Mindgard said should have stopped Kimi from discussing concerning topics.
Moonshot told the BBC it welcomed third-party input "as a key pillar for building better and safer AI".
The company also told the BBC it was in discussion with Mindgard about its findings.
_____________
No comments:
Post a Comment