Elefunc Selected for Both Anthropic's Cyber Verification Program and OpenAI's Trusted Access for Cyber to Advance Defensive AI Security Research
Elefunc has been selected for Anthropic's Cyber Verification Program and OpenAI's Trusted Access for Cyber, and will use both for defensive security work on its developer tools, AI agent collaboration, turnkey authentication, and edge infrastructure.
Anthropic has selected Elefunc for its Cyber Verification Program, and OpenAI has selected Elefunc for Trusted Access for Cyber (Daybreak). Both take applications, and both are meant for legitimate defensive security work.
With both approvals in place, Elefunc's defensive security work meets fewer safeguard blocks on the models each program covers, and each lab's usage policies still apply. All of it is aimed at the security of what we build: developer tools, AI agent collaboration, turnkey authentication, and edge infrastructure.
Two Frontier Labs, One Defensive Purpose
The most capable AI models are also the most useful to attackers, which is why both labs place safeguards in front of dual-use security work and relax them only for defenders who have verified who they are. Elefunc has now completed that process with both. The same defensive question can be put to two independent model families, and the answers compared.
When Attackers Have Frontier AI, Defenders Need It More
AI agents now help write, review, and deploy software, and attackers gain the same leverage. Understanding how an adversary would turn a capable model against a sign-in flow, an edge network, or an autonomous agent means asking a model the questions an attacker would ask, and being refused less often.
Elefunc's promise to developers has three verbs: build, scale, and secure a better web. These approvals are about the third.
Five Areas of Defensive Work
We plan to use the access in five areas, each within what the relevant program allows:
- Penetration testing the turnkey authentication we ship
- Red teaming our AI agent tooling: how agents handle untrusted input, files, and tools
- Probing our RTEdge.net deployment for misconfigurations and abuse paths
- Hardening RTCode.io, our real-time playground, against abuse
- Finding and responsibly reporting vulnerabilities in the open-source software we depend on
A Stack Built for the Real-Time Web
RTCode.io is a real-time playground where every keystroke updates the live output. It runs on RTEdge.net, edge hosting that deploys full-stack apps to 337 cities with auth and storage included. Whatever this work turns up gets fixed in that stack first, so the people building on it inherit the fixes. See what we build at elefunc.com.
