Frontier Labs' Safety Spend Just 0.1% of Capex, Analysis Claims
Across multiple posts, AlexTensor cites industry statistics to argue that frontier AI labs' safety spending is drastically mismatched with their public rhetoric: while OpenAI and Anthropic call for pausing AI development, joint international governance, and warn of catastrophic civilizational risk, their actual resource allocation suggests safety is far from a priority.
Confirmed
- Per the cited statistics, only about 3.4% of combined employees at frontier labs work on safety.
- Safety research spending is about $280 million, roughly 0.1% of combined capital expenditure, versus about $360 billion for infrastructure. In another post, AlexTensor summarizes the two companies' internal allocation to AI safety and cybersecurity as under 1% of capex.
- Industry comparison: nuclear spends roughly 30-40% (or 32%) of capital costs on safety compliance, aviation about 15% of revenue, pharma about 25%—all far above AI labs.
Unconfirmed
- The figures come from industry statistics relayed by AlexTensor; the posts do not cite the original source or methodology, so accuracy remains unverified.
Why it matters
- The analysis directly challenges the credibility of labs' 'civilizational risk' narrative: if the risk were as claimed, safety investment should be far higher.
- For policy debates, calls for international governance and R&D pauses require proportionate internal safety spending to be credible.
2026-09-14 ~ 2026-09-16 · 6 related posts
Primary sources
- AI labs spend ~0.1% of combined capex on safety, versus 30-40% for nuclear power — AlexTensor · 2026-09-14
- Frontier Labs Spend Just ~0.1% of Capex on Safety, Analysis Finds — AlexTensor · 2026-09-15
- [source] Anthropic and OpenAI spend under 1% of capex on cybersecurity and AI safety, critic says — AlexTensor · 2026-09-15
- [source] Anthropic and OpenAI spend only ~1% of capex on AI safety, analysis shows — AlexTensor · 2026-09-16
2 near-duplicate retellings: AlexTensor · AlexTensor