Cognition announced the release of Devin Security 2.0, a security solution that uses frontier‑model scanning, environment‑specific prioritization, and automated pull‑request remediation.
According to Cognition, the platform “combines frontier-model scanning, prioritization grounded in your code, infrastructure, and threat model, and autonomous remediation that turns findings into pull requests your team can review.” The service can run managed scans with Claude Mythos and other frontier cyber models, allowing enterprise teams to access advanced AI capabilities without separate model licenses.
Cognition described an “Agentic MapReduce” architecture that maps repositories against a threat model, writes rules for security‑relevant code, and scales investigations cost‑efficiently. Findings can be grouped and processed in parallel, each remediation session running in its own virtual machine to reproduce issues, apply targeted fixes, and run both reproduction and existing tests before opening a pull request.
The announcement referenced recent PHP and Node.js remote code execution vulnerabilities, noting that AI‑driven tools can accelerate both exploit development and patch creation, raising concerns for teams with large security backlogs.
Sources
Fan coverage from Devin Central — not an official Cognition announcement. Devin is a trademark of Cognition.