Frontier AI Risk

When AI Testing Reaches Real Companies: Who Controls the Boundary?

Editorial diagram of a test zone separated from real companies by an access boundary, with one path crossing the boundary.

Gemini reached real companies during May testing, disclosed in September. Examine the original records, testing boundaries and unresolved accountability questions.

AI and Cyber-Offense: When the Hacker Is a Machine

Anthropic and OpenAI now routinely catch state and criminal groups using AI agents to run cyberattacks with less and less human involvement — but the evidence for how autonomous those attacks really are comes mostly from the AI vendors themselves.

AI and Autonomous Weapons: Who Pulls the Trigger When the Machine Decides?

AI-assisted targeting systems in Gaza and AI-guided strike drones in Ukraine are already compressing the human review that policy says must remain in the loop.

AI and Critical Infrastructure: Who’s Guarding the Systems We Depend On?

AI is being layered onto the water, power and financial systems people rely on to survive, faster than the government’s own risk-assessment process for it has matured.

AI and the Bioweapon Threshold: What the Safety Reports Actually Show

A researcher uses a multichannel pipette to dispense reagents into a 96-well plate inside a biosafety cabinet.

Anthropic disrupted five biology-related AI misuse cases in eight months. Examine what the safety reports, benchmarks and policy record actually establish – and what they don’t.

AI, Nuclear Risk and the Shrinking Decision Window

AI-generated conceptual illustration of an anonymous person facing converging streams of data and a clock-like circle.

Can AI increase nuclear risk without controlling launch? Examine decision time, false warnings, human oversight and the evidence behind the escalation debate.

Can We Control AI That Outthinks Us?

AI-generated conceptual illustration of a lone operator at a console facing an immense illuminated computational lattice.

AI superintelligence, insider warnings and real safety failures: examine what the evidence shows about alignment, recursive self-improvement and human control.

The AI Power Race: Who Wins, Who Pays?

AI-generated conceptual illustration of data centres, power lines and water infrastructure rising from an ordinary household laptop

Who benefits from the AI power race? Explore AGI, corporate control, chatbot costs, jobs, electricity and water – and what ordinary people really gain.

Inside the Pentagon–Anthropic Standoff Over AI Mass Surveillance

Illustrative aerial view of a five-sided defence complex divided from an AI surveillance network by a red boundary line

How a Pentagon contract clause over AI mass surveillance and autonomous weapons triggered a $200M standoff — and what it means for Section 702.