'The walls are paper': AI safety debaters warn frontier models outpace global infrastructure
Justin_Halford_ · x · 2026-09-19
A debate on AI and cybersecurity: the author concedes that at steady state the world will formally verify software and build a permissioned internet, but argues there's a massive overhang between frontier models (especially with safeguards stripped) and the rest of the world's digital infrastructure — "the walls are paper." He cites an alleged OpenAI hack by three people using Claude this week, warns infra and compute will be hacked in reality, and that "millions could die."
More from Safety
- Guardian readers debate: existential AI warnings or a sleight of hand? — nordicinst · 2026-09-19
- Critics say OpenAI's eval environments aren't airgapped and barely monitored — MoonL88537 · 2026-09-19
- Anthropic's lobbying spend surges: 9 firms and $920K external, $1.97M in-house — kevinnbass · 2026-09-19
- Virginia, world's largest data center hub, unveils new restrictions on large projects — Polymarket · 2026-09-19
- Sam Altman confirms no OpenAI IPO in 2026, calling it an 'ill-advised moment' on safety — fortune · 2026-09-19
- OpenAI hacked by white hats who reported the bugs, rewarded with just $6,500 bounty — EconomySerious · 2026-09-19