Cognition opens Devin Security scans on Claude Mythos and pledges hundreds of millions in credits
Monday’s post: deep scans in Devin Security now run on Claude Mythos and other frontier cyber models, without a separate contract for those models. Current and future enterprise customers are offered free scans. New customers can also get remediation credits. Cognition does not say how the credits are split, or which Mythos version is in the scan.
Short2 minEasy
Priya Shah2 min read
- Devin
- Cognition
- Claude Mythos
- Security

Photo: CognitionSource page
Cognition said on Monday that Devin Security can now run deep scans with Claude Mythos and other frontier cyber models, and that enterprise teams can use those scans through Devin without separate access to the models. The post calls Devin Security model-neutral: one product, with the lab choosing which cyber model fits a job as models change. It also says Cognition is committing hundreds of millions of dollars in Devin credits. Current and future enterprise customers are offered free scans with those models. New customers can also receive remediation credits. The post does not say how many credits a customer gets, or what the product costs when the credits run out. It does not name a Mythos version.
The scan is aimed at what Cognition calls complex chained exploits. Findings from Devin’s own scans, and from scanners, bug bounties, and advisories that Devin ingests, are scored against a threat model the company says it builds with the customer’s security and engineering teams. That document is supposed to record services, network rules, what is deployed, and what the customer says would happen if one user could read another’s data or skip payment. Resolving a finding, the post says, is not always a code change. Devin can document a false positive or propose a risk adjustment, and each of those proposals requires human approval. The original CVSS score stays beside any adjustment. Vulnerabilities on the Known Exploited Vulnerabilities list are never relabeled. With access configured, Devin can also apply a firewall rule or close a public service while a fix is prepared. The underlying flaw stays open.
The fix path is a pull request, not a silent edit. Cognition says Devin can run thousands of parallel sessions, each in its own virtual machine, reproducing a vulnerability, writing a change, running the reproduction test and existing tests, and opening a pull request with evidence. A broader refactor, such as moving authorization checks into one place, is described as an estimate and an implementation the product can help with. What Monday’s post establishes is a managed way to point restricted cyber models at a customer’s code, and a credit pool whose size is stated only as hundreds of millions of dollars. What it does not establish is a measured rate of real vulnerabilities found, or of fixes merged.


