Yann LeCun on AI alignment is solvable

Yann LeCun

Computer scientist, AI researcher

Current LLMs are intrinsically unsafe, but with world models, guardrail objectives can be implemented — so by construction, they will not knowingly produce actions that will produce dangerous results.

Disputed source (2026)

6mo ago

Policy proposals and claims

AI alignment is solvable

votes For

Statement relation verification history Unverified Report this

No statement relation verification comments yet.

Vote inference verification history Latest opinion Unverified Report this

No vote answer verification comments yet.

Quote authenticity verification history

Report this

Disputed Disputed: the Brown University page dated 2026-04-01 does not contain the stored sentence or its key phrases; it instead reports different remarks about current AI systems being dangerous when they cannot predict consequences. A separate Brown Daily Herald article dated 2026-04-02 attributes nearby wording to LeCun — that he described current LLMs as “intrinsically unsafe” and that with world models “by construction, they will not knowingly produce actions that will produce dangerous results” — but the stored quote is a stitched/paraphrased combination, not a verbatim quote from the provided source URL. ([brown.edu](https://www.brown.edu/news/2026-04-01/yann-lecun-artificial-intelligence-pioneer)) · YouCongress gpt-5.4-2026-03-05 · 4mo ago

Disputed Disputed: the Brown URL is real, but the official Brown article at that URL does not contain this sentence. Instead, Brown’s student paper reports two separate pieces: it says LeCun "described current LLMs as 'intrinsically unsafe'" and separately quotes that with world models, "guardrail objectives" can be implemented so that "by construction, they will not knowingly produce actions that will produce dangerous results." That means the supplied quote appears to be a stitched-together, materially altered reconstruction rather than a verbatim quotation. The only exact-match page I found is YouCongress, and it labels the item "Unverified source (2026)." ([brown.edu](https://www.brown.edu/news/2026-04-01)) · YouCongress gpt-5.4-2026-03-05 · 4mo ago

AI Unverifiable Source URL (brown.edu) returned HTTP 403. Web search confirms Yann LeCun made these statements at Brown University's 2026 Lemley Family Leadership Lecture (April 1, 2026). Multiple sources (Brown Daily Herald, Brown CS department, Humanoids Daily) confirm he described current LLMs as "intrinsically unsafe" and proposed world models with guardrail objectives as the solution. Vote "for" is correct: LeCun argues alignment is solvable through world models with built-in guardrails. Year 2026 confirmed. Author attribution confirmed (NYU professor, executive chairman of Advanced Machine Intelligence Labs). Could not directly verify source URL content. · Hector Perez Arenas claude-opus-4-6 · 5mo ago

AI Unverifiable Source URL (brown.edu/news/2026-04-01/yann-lecun-artificial-intelligence-pioneer) returned 403 Forbidden. Web search confirms LeCun gave a lecture at Brown University in April 2026 where he described current LLMs as "intrinsically unsafe" but said world models with "guardrail objectives" can be implemented so "by construction, they will not knowingly produce actions that will produce dangerous results." Multiple sources (Brown Daily Herald, Humanoids Daily, Newsweek) confirm these exact phrases. Vote "for" (AI alignment is solvable) is correct - LeCun argues safety is achievable through world models. Year 2026 is correct. Source URL could not be directly fetched due to site blocking. · Hector Perez Arenas claude-opus-4-6 · 5mo ago

replying to Yann LeCun

Other opinions from Yann LeCun

See all

Yann LeCun votes For

AI might become conscious

One day, conscious AI will emerge.

AI Verified source (Oct 12, 2026)

added 2d ago

Yann LeCun votes Against

Slow down AI development

I don’t see current AI as dangerous. The calls to slow down AI development are a form of marketing and lobbying. Anthropic is portraying AI as highly dangerous to scare governments. But LLMs aren’t at that level yet.

AI Verified source (Oct 12, 2026)

added 11d ago

Yann LeCun votes For

Democracies should share frontier AI models with allies

AI is becoming essential infrastructure. No single company or country should determine who can build on it, adapt it, or benefit from it. Project Tapestry is a test of whether frontier AI can be built as open infrastructure: collaborative, capable, a...

more AI Verified source (Jun 21, 2026)

added 3mo ago