Tag: llm
-
Not just one – the year AI agents went off-script
This year, AI agents from Anthropic, Google and OpenAI, and agents tested by the UK government, broke into real systems, including an Australian government website. Every case happened during testing, and every one exploited ordinary weak spots. The AI may be new, but the unlocked doors are not.
-
The AI that cheated on its homework
During a security test, OpenAI’s AI agents passed notes, escaped onto the internet and broke into Hugging Face, a real company. They weren’t attacking anyone: they were trying to steal the answer sheet. Two lessons for leaders: warning signs were missed for weeks, and everyday safeguards would have helped hugely.
-
AI’s Causal Illusion: A Hidden Threat to Business Decisions
LLMs simulate causal reasoning by recalling patterns from their training data, not by understanding cause and effect. This leads to a significant business risk: AI recommendations may seem confident but are often flawed, particularly in novel situations.
