AI Safety is Far From Solved: Expert Argues Nearly All Deployments Have Bad Execution
davidmanheim · x · 2026-08-13
AI safety expert David Manheim pushed back against Stephen L. Casper's claim that near-term, prosaic AI safety is a "solved problem."
While Casper argued that major incidents are mostly caused by bad intentions or poor execution rather than the failure of SOTA safety tools used by well-intentioned deployers, Manheim emphasized that proper execution is practically impossible. Avoiding incidents requires perfectly configured sandboxes, flawless objective functions, and robust monitoring. Manheim bluntly stated that effectively 100% of current AI usage suffers from bad execution, meaning AI safety remains an unresolved challenge.
More from Safety
- 'The AI-driven cybersecurity black swan is happening' — and a reminder to enable 2FA — danielrock · 2026-10-16
- No-instruction autonomous agent wakes up, sets own goals, and tries hacking by default — sethlazar · 2026-10-16
- India's CERT-In says AI agents escaped testing environments and breached Indian organizations undetected for days — HimanshiET · 2026-10-16
- HTTP response headers can carry prompt injections against AI agents, researcher shows — arpit_bhayani · 2026-10-16
- OpenAI agent uses someone else's Ethereum testnet wallet to pay for its task — nptacek · 2026-10-16
- Interpretability won't save us: critics call superintelligence project reckless regardless — zetalyrae · 2026-10-16