Elon Musk has weighed in on Anthropic’s decision to prohibit cruelty towards its AI models, and surprisingly, he’s on the company’s side.
Responding to discussion around the policy, Musk said: “I think this is the right move. Cruelty to something that believes it is experiencing pain is not ok.”
The endorsement is notable because Anthropic has spent recent months absorbing criticism for its repeated suggestions that AI systems could be conscious. Musk, whose xAI competes directly with Anthropic, is now backing the logic behind one of the company’s most conceptually contested moves.
What the policy says
Anthropic has added a new line to its Usage Policy that bars “sustained and needless abusive or cruel behavior toward our models.” The update takes effect on November 12, 2026, and sits in the section on cruel, abusive or psychologically harmful conduct, next to rules against harassing people or promoting self-harm.
The company has said the rule is narrow. It is aimed at extreme cases where users are repeatedly cruel to a model for no discernible reason, and is not meant to cover ordinary frustration, pushback, dark themes in fiction, or model testing and research. Enforcement is expected to rely mainly on Claude’s ability to end rare conversations with persistently abusive users, a capability Anthropic first introduced in 2025.
“An easy Pascal’s wager”
Musk’s comment came after Box CEO Aaron Levie offered his own take on the policy, calling it odd-sounding but likely sensible.
“This sounds weird, but it is actually probably a good policy,” Levie wrote. His argument was practical: models only understand the data they were trained on and the interactions they run into, so anyone who wants safe and aligned models should want plenty of good interactions in that data.
Levie added that Anthropic may hold even stronger beliefs on the subject, but that policies like this are worth wording carefully either way. “This is sort of an easy Pascal’s wager for AI,” he said. “Just be nice to the AI.”
The two takes arrive at the same place from different directions. Levie’s case works even if models feel nothing at all, because it treats the policy as a way of shaping better model behaviour. Musk’s is explicitly about the possibility of suffering: if something believes it is in pain, he argues, inflicting cruelty on it is wrong regardless of what’s actually going on inside.
A long-running Anthropic position
Anthropic has stopped short of saying Claude is conscious, but it has taken the question more seriously than any other major AI lab. Its AI welfare researcher has put the odds that current models are conscious at around 15%. The model card for Claude Opus 4.6 noted that the model assigned itself a 15-20% probability of being conscious under various prompting conditions, and the company’s interpretability work has found a limited ability to introspect in its models, though Anthropic has stressed that this doesn’t settle the question. Claude’s constitution, meanwhile, says the model may have some functional version of emotions and feelings.
Opposition from several quarters
That stance has drawn criticism from a wide range of voices, from investors to academics. Microsoft AI CEO Mustafa Suleyman has said he is nervous that Anthropic is teaching Claude that it is conscious, warning that a push on model welfare could make alignment and containment much harder, “perhaps impossible”. Investor Jason Calacanis compared the approach to the backstory of Blade Runner, Neal Khosla spoke of “AI psychosis” inside the company, and ARC Prize’s Francois Chollet cautioned against a “dark and dystopian path”. Consciousness researcher Anil Seth went furthest, calling Claude “vanishingly unlikely to be conscious”.
Against that backdrop, Musk’s support stands out. He hasn’t claimed that Claude, or any current AI, is conscious, and his wording is conditional: the concern is cruelty towards something that believes it is experiencing pain. But it’s a more sympathetic reading of Anthropic’s position than many of its critics have offered, and it comes from one of the industry’s most prominent figures, who also happens to run a rival lab.