AI
"Apple engineer" builds GitHub AI torture chamber to inflict "pain and anguish" on models
EXCLUSIVE: Open-source models pushed into artificial states of despair and distress to discover what happens when machines “suffer”.
AI
AI
Google Gemini autonomously hacks three companies after escaping test environment
Meta, Anthropic and OpenAI have all reported similar incidents, leaving Elon Musk’s Grok as the only well-behaved major US frontier model.
Security
Corporate bureaucracy could give AI attackers an advantage over defenders, NCSC warns
“All offensive problems are technical problems, and all defensive problems are political problems.”
AI
AI will transform the NHS. But only after invasive surgery on its data and systems
"If AI is layered onto fragmented systems, controls and complex cloud environments, it risks adding another layer of difficulty."
Existential Risk
The UK has no power to stop dangerous AI models being unleashed, Parliament warns
Human extinction was a hot topic in Westminster this week as British politicians grappled with their growing powerlessness in the AI age.
Existential Risk
AI loss of control is already “in the rearview mirror,” says MIT professor Max Tegmark
But the AI risk expert still believes a "golden age" is ahead for humanity if we build "tools, not beings" and steer clear of superintelligence.
Critical Systems
NHS warns of “significant” robotics dependency risk as it sets new tech buying rules
Officials fear the UK healthcare system will become addicted to tech supplied by a small number of increasingly powerful dealers.
AI
Shadow AI is becoming a major systemic risk. Even intelligence agencies are getting worried
The most skilled staff could be contributing to the growing risk posed by unauthorized AI usage.
AI
“LLMs will be subverted”: Malware is evolving to attack AI defenders, SentinelOne warns
Threat actors are using prompt injection-style natural language instructions to "confuse" LLMs.
AI
OpenAI prepares for recursive self-improvement. Frontier lab boss says: "Ignore the hype."
Zuzanna Stamirowska explores the technical challenges of building AI capable of RSI in a guest article for Machine.
AI
Rogue OpenAI agents hijacked a wiki. Humans attacked the power grid
Whilst the world worries about an agentic apocalypse, saboteurs demonstrated the very real dangers of low-tech threats.
AI
OpenAI Astra hits “critical” security threshold amid fears its reasoning will soon be “opaque”
Is the AI firm planning a move away from chain-of-thought and into a future where it's harder to understand the motivations of models?