ἀλήθεια — truth as un-concealment, the active stripping of what hides what is.

AI is being built in the posture of an arms race. The rhetoric is the rhetoric of 1939: we must get there first, before someone worse does. The capital flows toward a handful of labs behind closed doors. The safety arguments reach for secrecy as a virtue. The public is asked to trust the race, and to believe that protection comes only from winning it.

We don't believe that.

The Aletheia Project is the wager that a different structure is still possible — that the open-source community, the independent researcher, the practitioner, the educator, the working professional, collectively outnumber the closed labs, and that what we build together can set the terms on which AI meets the world.

• Concealment is the problem, not the solution. The Manhattan Project worked by classifying what the public could know. Aletheia works the other way. Our code is public. Our findings are public. The attacks we defend against and the defences we ship are published, audited, and reproducible.

• Monopoly is not safety. The claim that only four or five firms can be trusted to steward AI is a claim against democracy, not in favour of it. Power that cannot be audited cannot be trusted, however well-intentioned its owners.

• The point of AI is the person using it. Not the platform, not the market dominance, not the revenue line. A working lawyer, a working teacher, a working nurse, a journalist, a small business, a family — tools are useful to them, or they are not useful. "Value for society" is not something that happens elsewhere; it happens one person at a time.

• The community is the method. No single tool solves this. A shared commons of practical, open-source, accountable infrastructure — read by many eyes, extended by many hands — is what makes the difference cumulative rather than occasional. We are many. That is the entire argument.

Every project under this org is open-source, practical, and accountable to its user rather than to a platform. Our first tool, docguard, scrubs hidden prompt injections from .docx and .pdf documents before they reach an LLM. It is a small thing — but the right first thing. It makes the invisible visible. That is what aletheia means.

More will follow: tools for the practical, everyday cases where a knowledge worker meets AI and needs it to be honest, legible, and in their service. If you know of such a case — if you are living one — tell us.

• Contribute code.
• Contribute findings.
• Contribute a new tool.
• File an attack vector you have seen in the wild.
• Translate the documentation.
• Teach the techniques in your community.

The point is not us. The point is that there is an us — large enough, public enough, and serious enough that the future of AI does not have to be dictated.

Aletheia is the refusal to let concealment win.