Further findings: The full grant is reported paid, and a third contract commissions evaluation design.
New: we obtained a UK contract. Read the publication and commercial-reuse terms, including corrected award values and a separate recorded payment.
The useful discovery in Upper Echelon’s September 17 video is the evaluator itself. Irregular sits between the people supplying capital, the labs building models and the institutions deciding what counts as adequate AI security. Following its public records adds a substantial branch to the EA network we have been mapping.
Open the funding and people map, or compare the incident accounts.
Philanthropic money before the venture announcement
A preserved Good Ventures page lists a February 2024 grant of $6.8 million over two years to Pattern Labs, recommended by Open Philanthropy, for security-related software and analysis. The record describes a grant, not an equity investment. Good Ventures and Open Philanthropy occupy different roles in that one award.
The company’s later launch announcement explicitly identifies Irregular as the former Pattern Labs and announces $80 million raised with Sequoia and Redpoint. The name change is publicly acknowledged. The earlier grant’s old live URL now redirects to Good Ventures’ giving overview; the historical entry is available in the archive.
Sequoia’s investment account explains the commercial proposition unusually clearly: close work with Anthropic, OpenAI and Google DeepMind; pre-release security testing; defenses; and compliance standards and roadmaps. This is a business selling expertise in a field where that expertise can help shape the rules.
The people connect the institutions
Omer Nevo is Irregular’s co-founder and CTO. EA Israel lists him on its board. Probably Good identifies him as its co-founder, former CEO and current director. Heron lists him as an adviser, names EA Israel as its institutional host and discloses Coefficient Giving funding. Heron helps cybersecurity practitioners find opportunities in AI security.
Dan Lahav, Irregular’s co-founder and CEO, co-founded Impact Focused Education with Sella Nevo. Sella also co-founded Probably Good and remains a director. His current RAND role is founding director of the Center on AI, Security, and Technology. Older partner biographies still describe the Meselson Center role.
The education funding is specific too. EA Funds’ ledger, which returns CSV, lists one joint $394,968 grant to Dan Lahav and Sella Nevo in the third quarter of 2022 for an online course about pursuing high-impact work. It does not identify the course or receiving organization, so the map keeps the grant attached to its named recipients.
Lahav and Sella Nevo also co-authored RAND’s 2024 Securing AI Model Weights. The connection is professional work with a published output, reaching beyond shared affiliations.
What this adds to the bigger picture
Our reading is that this branch joins three kinds of capacity: philanthropic support for a security project, commercial investment in an evaluator, and institutions developing people for adjacent work. The same professional network participates in education, technical services and policy-facing research. That is a stronger and more useful finding than treating every EA connection as interchangeable.
It also suggests a practical question: when a testing supplier becomes important to several competing labs, how readily can those labs replace it or independently challenge its methods? The next documents with the most value are evaluation contracts, scope approvals, containment checks, notification records, liability terms and the governance of the proposed industry guidance.
This pass follows the primary records reached through the video and Effort’s investigation. The video and full automatic captions are preserved in our private video archive. The public database contains the relationships, source links and this analysis.