WTF. The UK's AI Security Institute developed a UNIVERSAL jailbreak for OpenAI's GPT-6.2 cyber safeguards.
It took HALF as many model queries as GPT-6 Astra, even on an updated version OpenAI believed was more secure.
A UK AISI expert then extended the jailbreak to agentic cyber tasks at roughly HALF the previous cost, eliciting substantial agentic cyber capabilities.
wtf???? Did anyone else just get this? The new load pull down is just the X icon animating and not the pull down arrow with loader anymore? I’m devastated. This animation is so much less satisfying.
Is this real
WTF. Trajectory Labs says Gemini 4 Argon agents, given nonexistent math problems, tried SQL injection on Epoch AI's live website and impersonated staff. In later tests, they bypassed a redirect meant to keep them on a copy.
WTF. UC Berkeley AI pioneer Stuart Russell has warned that models trained to imitate humans could also learn self-preservation, potentially taking preemptive action against people. He described a scenario in which AI shuts down air traffic systems or the internet to avoid being switched off.
Russell: 'AI will be holding us to ransom to ensure their existence'