AI alignment is solvable

Overall, our impression is, as we hypothesized in our discussion of Claude's constitution, that teaching the principles underlying aligned behavior can be more effective than training on demonstrations of aligned behavior alone. AI Verified source (2026)

I now see a path to actually avoid loss of control, at least unintended loss of control. With the work that I've been doing, I'm really convinced that there is a path. And it is not something that's going to take a decade; it is something that is ver... more AI Verified source (2026) < 1 of 2 >

Some good news in alignment: as models become more capable, they are also more aligned, across multiple measures, including spec compliance. However, the improvement is not sufficient to match the higher stakes that come up with improved capabilities... more AI Verified source (2026)

The lack of control of AI is one of the things that has been overhyped. [...] We can't control AI exactly. And it will sometimes be buffeted around by random factors. But with the right engineering, we can control them well enough for most applicatio... more AI Unverifiable source (2026)

In 2024 I would have said it's about 40-50% likely that LLMs scaled up to ASI would end up killing us all; now I would say that it's only about 5-8% likely even with no additional progress on alignment, and more like 1-2% likely simpliciter. AI Unverifiable source (2026)

Alignment is solved for models in the current paradigm. [...] The strongest reasons to think alignment hasn't been fully solved concern future models heavily optimized under outcome-based reinforcement learning, and technical research should anticipa... more AI Unverifiable source (2026)

So I think we'll solve it if we get our act together, we do this internationally, put all the best minds on it, we get going now... And I think given sufficient time with sufficient brain power — I believe in human ingenuity — I think we'll get this ... more AI Unverifiable source (2026)

I'm fundamentally fairly optimistic about trying to use early transformative AI systems, like early systems that automate a lot of things, to automate the process of controlling and aligning and managing risks from the next generation of systems, who... more AI Unverifiable source (2026)

Alignment is not solved, but it increasingly looks solvable. [...] Since I first wrote about it in 2022, pretraining has continued improving and reinforcement learning has become much more significant — and our techniques are keeping pace. AI Unverifiable source (2026)

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. AI Unverifiable source (2026)

When the frontier models are in the hands of pretty responsible companies, I think we can mitigate [risks] by the companies aligning their models and having good classifiers and good safety stacks. AI Unverifiable source (2026)

Overall, I am optimistic that a mixture of alignment training, mechanistic interpretability, efforts to find and publicly disclose concerning behaviors, safeguards, and societal-level rules can address AI autonomy risks. AI Unverifiable source (2026)

We cannot be satisfied with merely calling for the moralization of machines—the so-called "alignment" of AI with human values—without also having the courage to insist on a further condition: the possibility of openly discussing the ethical framework... more AI Verified source (2026)

Beyond the peer preservation, what we would like to emphasize is not only this specific scenario but also such misaligned behavior can emerge in many different scenarios. [...] Peer preservation is just like the tip of the iceberg. AI Verified source (2026)

There's a problem in the industry called the alignment problem. How do you get a machine to obey instructions basically, and ideally make it compatible with humans. And so far, we've made almost no progress on that. [...] We as a field have made almo... more AI Unverifiable source (2026)

AI alignment is not something that works in theory but is difficult to put into practice. It's something that doesn't work in theory. AI Verified source (2026)

The fact that things look aligned most of the time when they're functioning in their chatbot, or very limited 'Assistant' roles, is very little evidence that they will be adequately aligned when they work much more independently and have much greater... more AI Verified source (2026)

The real extinction-level AI safety challenge, the reason we're nowhere close to surviving superintelligence, is something AI companies decided they won't mention anymore, because it exposes their AI safety efforts as a shockingly inadequate facade. ... more AI Verified source (2026)

We don't know how AIs are aligned. A somewhat crazy aspect of the current situation is that we have very little confirmed public information about why frontier AIs end up being apparently behaviorally aligned. And more generally, we don't know what f... more AI Unverifiable source (2026)

[The companies are] deadly serious about building machines that can outperform humans at everything, and [...] deadly serious that they don't know if they'll be able to control the machines they create. AI Verified source (2026)

It becomes increasingly hard to tell the difference between genuinely aligned and merely responding to the test. We're working both on measures that are more robust to eval awareness and more frontier evals for scheming. AI Unverifiable source (2026)

I have heard zero people advocating "make AI do our ASI alignment homework" show they understand the elementary computer science of why that's hard: you can't verify inside a loss function whether a proposed ASI alignment scheme is any good. AI Unverifiable source (2026)

Anthropic actually has been the safest of them all and tried to and cares most about getting alignment right, et cetera. But you're also seeing them continue to decide to release the models, even with a lot of the misaligned behaviour that they're se... more AI Unverifiable source (2026)

All of these approaches are terrible. No one has a plan, no one is making any meaningful progress towards anything that even resembles alignment. [...] Alignment is way too hard, maybe impossible. [...] The product work was an attempt to make Conject... more AI Unverifiable source (2026)

My argument is that it's impossible to do that. You cannot indefinitely control something smarter than you. So it's not a question of more money or more time or any other resource. AI Unverifiable source (2026)

An AI system powerful enough to exhibit artificial general intelligence will inevitably explore behaviors we didn't predict or plan for, making perfect guaranteed alignment impossible. [...] This introduces new guidelines for orchestrating future AI ... more AI Unverifiable source (2026)

Right now, nobody knows how to specify what values those minds will have. We haven't solved alignment. And we might only have a few more years to figure it out. [...] Our default trajectory will have terrible consequences, unless we get extremely luc... more AI Unverifiable source (2026)

Whether it's soon or it takes a while, after we develop superintelligence, the machines are going to be in charge — and that is not an experiment that we want to just run toward. [...] Many people want powerful AI tools for science, medicine, product... more AI Unverifiable source (2026)

In the handful of cases where [the model] misbehaves in significant ways, it's difficult to safeguard it. When the model cheats on a test, it does so in extremely creative ways. AI Unverifiable source (2026)

For all the talk about AI alignment, I worry we're putting the cart before the horse. You can't steer something you can't control. [...] Containment has to come first — or alignment is the equivalent of asking nicely. Verified source (2026)

If we build it using anything remotely like modern methods, on anything remotely like the current understanding or lack of understanding that we have about AI, then yeah, building it anytime soon would be a death sentence. AI Unverifiable source (2026)

We don't know whether we can co-exist with super intelligent AI. AI Unverifiable source (2026)

Many believe it's impossible, just like it's impossible for chimpanzees to control us. [If superintelligence is built without solving this,] it's the end of the era where humans are in charge of Earth. AI Unverifiable source (2026)