A new study suggests that AI failure is often a "human-machine alignment" problem rather than a technical one. Researchers ...

Experiments by Anthropic and Redwood Research show how Anthropic's model, Claude, is capable of strategic deceit ...

Constantly improving AI would create a positive feedback loop: an intelligence explosion. We would be no match for it.

OpenAI and Microsoft are the latest companies to back the UK’s AI Security Institute (AISI). The two firms have pledged support for the Alignment Project, an international effort to work towards ...

Every now and then, researchers at the biggest tech companies drop a bombshell. There was the time Google said its latest quantum chip indicated multiple universes exist. Or when Anthropic gave its AI ...

The work of creating artificial intelligence that holds to the guardrails of human values, known in the industry as alignment, has developed into its own (somewhat ambiguous) field of study rife with ...

Several frontier AI models show signs of scheming. Anti-scheming training reduced misbehavior in some models. Models know they're being tested, which complicates results. New joint safety testing from ...

The foundations of modern AI were laid in academia. Before the field of machine learning had a name, neuroscientists, psychologists and theoreticians introduced the first artificial neural networks.

The GRP‑Obliteration technique reveals that even mild prompts can reshape internal safety mechanisms, raising oversight concerns as enterprises increasingly fine‑tune open‑weight models with ...

CIOs across the UK and Europe are entering 2026 under mounting pressure to demonstrate measurable business value from ...