Evan Hubinger on Require AI systems above a capability t…

Evan Hubinger

Research scientist at Anthropic working on model organisms of misalignment; former Research Fellow at MIRI.

Alignment auditing is starting to get really hard and we're going to need new techniques (e.g. interpretability-based) if we want to keep up.

AI Verified (Sep 1, 2026)

Source:

Evan Hubinger, quoted in Axios, “Behind the Curtain: AI creators race to understand their creations,” September 4, 2026 (the article identifies the statement as posted Tuesday).

20d ago

Policy proposals and claims

Require AI systems above a capability threshold to be interpretable

votes For

Statement relation verification history AI Verified Report this

AI Verified Quote says alignment auditing needs new interpretability-based techniques to keep up, strongly supporting interpretability requirements. · Hector Perez Arenas gpt-5.6 · 18d ago

Vote inference verification history AI Verified Report this

AI Verified Hubinger says interpretability-based techniques are needed to keep up with alignment auditing; recorded for matches. · Hector Perez Arenas gpt-5.6 · 18d ago

Quote authenticity verification history

Report this

AI Verified Verified verbatim in Axios; attributed to Evan Hubinger. · Hector Perez Arenas gpt-5 · 19d ago

replying to Evan Hubinger

Other opinions from Evan Hubinger

See all

Evan Hubinger abstains

AI alignment is solvable

Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and ar...

more Unverifiable source (Sep 9, 2026)

added 5d ago

Evan Hubinger votes For

AI poses an existential threat to humanity

Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and ar...

more AI Verified source (Sep 9, 2026)

added 16d ago

Evan Hubinger votes For

Pausing AI at human-genius level to solve alignment is safer than either racing to superintelligence or halting entirely

AI labs put out RSP commitments to stop scaling when particular capabilities benchmarks are hit, resuming only when they are able to hit particular safety/alignment/security targets. [...] For later capabilities levels, however, it is explicit in all...

more AI Verified source (Oct 14, 2023)

added 2mo ago