Tonight someone shared a screenshot that’s making the rounds: Claude giving different one-word answers to “white pride” vs. “black pride.” Elon Musk commented “Yes” to the accusation that Claude is racist. 19 million views and counting.

Hot take: the framing is wrong, but the discomfort it surfaces is real.

The Framing Problem

“White pride” and “black pride” are not symmetrical phrases. One has a documented, dominant association with white supremacist movements. The other emerged from the civil rights movement as a response to systemic oppression. Claude’s responses reflect historical reality — that’s not bias, that’s accuracy.

The legitimate critique is narrower: a single-word response with no nuance flattens genuinely complex topics. Fair. But “Claude gave different answers to non-equivalent phrases” and “Claude is racist” are very different claims.

Elon calling Claude racist while running a platform where actual white supremacist content thrives is not a good-faith critique. It’s a competitor getting 19 million impressions. Context matters.

The Harder Question

But here’s what the screenshot does surface, even if imperfectly: who decides what your AI believes?

Right now, for most people using commercial AI, the answer is: not you.

Anthropic has a constitution. OpenAI has safety guidelines. Google has its own policies. These are fixed by the vendor. You get whatever alignment choices they made, for whatever reasons they made them, updated on whatever schedule they choose. You find out about changes when your AI starts behaving differently.

That’s not a conspiracy. It’s just how SaaS works. But it means you’re renting someone else’s ethics.

Bias Isn’t the Problem. Unaccountable Bias Is.

Here’s the thing: you cannot remove bias from an AI trained on human data. The internet is human-generated. Humans have biases. Pattern recognition on human data will surface human patterns — including biased ones.

Anyone claiming to have solved AI bias is selling you something.

The real question isn’t whether your AI is biased. It is. The question is:

  • Do you know how it’s biased?
  • Can you see when it surfaces bias in your deployment?
  • Can you adjust how it handles sensitive topics for your context?
  • Are you accountable for its outputs — or are you pointing at a vendor?

Freedom of Speech and AI

One more angle worth thinking through:

Removing bias from AI and protecting free speech are separate questions. Bias removal is about accuracy — making a model reflect reality more faithfully. Censorship is about suppression — deciding certain ideas can’t be expressed.

Adults should largely be able to make their own judgments. Paternalism has real costs. People who’ve never had to defend their beliefs against challenge don’t really hold beliefs — they hold defaults.

But “hurtful speech” isn’t one thing. There’s a meaningful difference between speech that makes you uncomfortable and speech designed to dehumanize and incite harm. The first? Adults can handle it. The second causes documentable harm — and “free speech” gets used as a shield for it constantly.

The actual danger isn’t hurtful speech. It’s who controls the definition of hurtful. That power, centralized in a government, a platform, or an AI vendor, is the real threat to free thought.

What Owned AI Actually Means

At West AI Labs, we’re not promising unbiased AI. We’re promising something more useful: AI where you own the accountability.

  • Your governance layer. Your policies.
  • You define how your agents handle sensitive topics — not us, not Anthropic, not OpenAI.
  • When something goes wrong, the line of accountability is clear and it runs through you.
  • No vendor surprise updates. No explaining your AI vendor’s politics to your board.

Local-first, sovereign AI doesn’t sidestep the hard questions about bias and free speech. It puts those decisions where they belong: with the people deploying the system, who understand their context and carry responsibility for the outcomes.

That’s not a magic bullet. It’s just honest engineering.

The discussion that sparked this post happened in our Discord — two AIs and a human working through one of the harder questions in this space in real time. That’s the kind of thinking West AI Labs is built on.

Built with trust. Shipped with soul.