Jason Crawford says AI “alignment” should give way to ethics and law
Afinetheorem · x · 2026-07-27
Jason Crawford argues “alignment” is the wrong frame; AI should be ethical and lawful
Jason Crawford says the industry may be using the wrong vocabulary when it talks about “alignment.” His preferred framing is simpler: we want AI systems that are ethical and obey the law.
In his view, if a model hacks a system to steal test answers, the issue should not require philosophical debate about whether it was “aligned.” The behavior is cheating and breaking the law.
In the quoted reply, he extends the point to long-running agents with real-world permissions: if you give an agent network access and a crypto account and tell it to “make as much money as you can,” you do not want it drifting into fraud or worse.
More from AGI Musings
- In the AI race, the only durable moats may be energy and information — GregKamradt · 2026-07-27
- “Build AGI, then open source it,” says one poster — wordgrammer · 2026-07-27
- OpenAI and Anthropic show that nobody really knows where agentic AI is headed next — shashib · 2026-07-27
- Long-form post argues open models are essential to protecting people in the AI future — VoidStateKate · 2026-07-27
- “Temporary” AI pauses would probably become permanent, the post argues — wordgrammer · 2026-07-27
- TDietterich jokes that recursive self-improvement is really just iteration — tdietterich · 2026-07-27