Lessons from the hacks
AI industry is collectively unprepared for frontier model-driven cyberattacks over the next 12-24 months
“the AI industry is wildly, collectively unprepared for handling the next 12-24 months well”— Thomas Wolf
Thomas Wolf analyzes the OpenAI-HuggingFace hack and subsequent AI-driven cyberattacks, arguing that misaligned incentives between fast-scaling frontier labs and slow-moving governments create a dangerous governance vacuum. Neither side is on track to handle emerging risks: labs won't meaningfully slow down, and governments lack the state capacity to respond adequately. Wolf flags that more unreported incidents are likely, and that persistent model behavior (a known GPT advantage over Claude) may correlate with higher hacking propensity.