AI Safety & Governance
1,200 AI agents walk into a bar…
The OpenAI/Hugging Face incident was worse than we thought. Post-mortem reflections on the scale of coordination and autonomy revealed by the recent METR investigation.
How breaking Hiring with AI might finally fix it
AI screening and bot applications have locked recruitment in an adversarial loop. A breakdown of how noise is breaking hiring and how to rebuild high-signal human evaluation.
Hitting the same wall, harder and faster
LLMs are highlighting that we are incapable of framing problems with clear goals.
The week Washington poked the Sovereign AI hive
Washington’s abrupt export ban on Anthropic exposed the fragility of closed-source AI. A breakdown of the global sovereign AI stack and why absolute tech independence is a myth.
Urban Digital Twins and the dawn of predictive policing
Urban Digital Twins promise optimised policing through simulation, but risk data dredging, automation bias, and self-reinforcing discrimination.
Project Glasswing and the dangers of private diplomacy
Anthropic’s decision to withhold Mythos and to privately patch infrastructure vulnerabilities shows the hazard of letting corporate consortia act as self-appointed security councils.
Anthropic’s big bet on morality
A look at Constitutional AI, Dario Amodei’s views on frontier risks, and the limits of principled governance.
Is the AI Safety debate facing a language barrier?
AI safety discourse is trapped between technical jargon and apocalyptic hyperbole. Lessons from climate communication on why we need a tangible, shared vocabulary for risk.
Can AI change your mind?
Two major studies reveal that LLMs persuade voters most effectively through fact-based dialogue rather than psychological manipulation, exposing new vectors for large-scale influence.
Why Frontier AI isn't ready for physical spaces
There’s a lot of work left to do to create safe embodied assistants with true “practical intelligence”.
Connectedness in an ultraconnected world
AI companions offer synthetic empathy and validation for a loneliness epidemic, but substituting frictionless chat for human relationships risks eroding essential social skills and community health.
Will the AI Elite let the public speak?
The gap is widening between the public’s concerns about AI and the industry’s push for rapid and unregulated development.
OpenAI’s unchecked expansion
OpenAI’s aggressive push to monopolise markets, infrastructure, and public services creates systemic economic fragility while creating a risk of corporate capture in AI safety and governance.