Policy
152d agoAI Safety: Progress and Persistent Challenges
Across the world, AI systems are growing faster than ever, pushing the limits of alignment and safety. While leading labs such as OpenAI and Anthropic report better spec compliance, they still struggle with adversarial robustness, dishonesty, and reward hacking.
The Signal
International policy makers must coordinate to address these risks before next‑generation models emerge.