Anthropic says Claude found new weaknesses in cryptographic
Anthropic says Claude Mythos found faster ways to attack cryptographic algorithms, highlighting how AI could uncover hidden security weaknesses.
TL;DR
Anthropic upgraded Claude Security on October 21, 2026 to scan code continuously at each commit, automatically track and resolve vulnerabilities, and integrate findings via API and MCP. The change tightens security coverage and offers usage credits, potentially reshaping enterprise code‑security practices.
Key facts
- Scanning now triggers on every new commit instead of a calendar schedule.
- Projects can include up to eight repositories from GitHub or GitLab.
- Eligible customers may receive up to $25,000 in monthly credits for two months.
Anthropic announced on October 21, 2026 that its Claude Security tool now continuously scans connected code repositories, shifting from scheduled scans to commit‑triggered analysis. The public‑beta update, aimed at Claude Enterprise customers, automatically rescans each repository whenever a new commit lands and tracks every identified vulnerability until it is resolved. The change also introduces automatic reopening of issues if a previously fixed flaw reappears, and expands integration options through API and Model Context Protocol (MCP). The rollout includes up to eight repositories per project and a two‑month credit incentive for eligible organizations.
Background
Claude Security entered public beta in April 2026, offering organizations a way to leverage Anthropic’s Claude model family, including Claude Mythos, for automated code vulnerability detection. Since its debut, the service has been used by thousands of organizations and has reported identifying over one million vulnerabilities across those deployments. Anthropic has also pursued an open‑source focus, launching an opt‑in OSS Scanner that examined 25 open‑source projects and surfaced 493 issues. The original scanning approach relied on point‑in‑time scans that left gaps between runs, meaning code could be shipped without immediate analysis. User feedback highlighted the need for tighter integration with existing development workflows and more timely detection of regressions. In response, Anthropic rebuilt the monitoring engine to trigger scans on every commit, provide lighter incremental scans after an initial deep baseline, and automatically mark findings as fixed when the corresponding code change is detected. The tool also supports exporting findings via API and the emerging Model Context Protocol, a standard for linking AI systems to external software, to ease adoption within existing security stacks.
What happened
On October 21, 2026 Anthropic released an update to its Claude Security product, shifting the scanning cadence from calendar‑based runs to continuous monitoring triggered by new commits, as reported by Cryptobriefing. The update introduces an initial deep baseline scan when a repository is first connected, after which the system performs lighter scans focused solely on the changes introduced by each subsequent commit. Each vulnerability discovered is now tracked individually until it is marked resolved; the tool automatically detects the code change that fixes a flaw and marks the finding as fixed, and it can reopen the issue if the same vulnerability resurfaces later. The service allows projects to include up to eight repositories from platforms such as GitHub and GitLab, and findings can be exported through an API or via the Model Context Protocol (MCP) for integration with external tools. The public beta is available to Claude Enterprise customers, with eligible organizations able to receive up to $25,000 in monthly usage credits for two months, an offer that runs until December 31, 2026. Alongside the main product, Anthropic is testing a parallel beta plugin called Claude Code, which enables developers to run local, session‑specific scans on particular repositories or individual changes. The company’s track record includes more than one million vulnerabilities identified since the tool’s public beta launch and 493 issues uncovered across 25 open‑source projects through its OSS Scanner.
Market impact
The Claude Security update is positioned to influence security tooling adoption among enterprises that already use Claude Enterprise, as the continuous scanning model reduces the window of exposure between code changes and vulnerability detection. The two‑month credit incentive of up to $25,000 per month may accelerate uptake among organizations evaluating AI‑driven security solutions before the offer expires on December 31, 2026. By allowing findings to flow into existing security information and event management (SIEM) systems via API and MCP, the update could drive integration spending and potentially shift market share away from legacy static analysis tools that rely on manual scheduling. The eight‑repository limit per project may prompt larger firms to segment their codebases across multiple Claude Security projects, potentially increasing overall usage volume. While specific price movements or token flows are not detailed in the source, the announcement signals a broader trend of AI‑enhanced security services gaining traction in the crypto and broader tech sectors.
Why it matters
The shift to commit‑triggered scanning narrows the time gap during which new code could contain undiscovered flaws, addressing a key weakness of point‑in‑time scans that leave code unexamined until the next scheduled run. Automatic resolution tracking reduces manual verification effort for security teams, allowing them to focus on remediation rather than confirming fixes. The ability to automatically reopen issues if a vulnerability reappears adds a safety net against regression, a scenario that can be more damaging than an initially unknown bug. Integration through API and MCP means findings can be fed directly into existing security workflows, increasing the likelihood of adoption and reducing friction. The credit incentive lowers the financial barrier for early adopters, potentially expanding the user base and generating more data to further refine the model’s detection capabilities. Together, these changes could accelerate the maturation of AI‑driven code security and set new expectations for continuous, automated vulnerability management in the software development lifecycle.
DeFi Intel’s take
Editorial analysis, not a statement of fact.
DeFi Intel's take: the continuous‑monitoring upgrade suggests Anthropic is positioning Claude Security as a direct competitor to traditional static analysis tools by eliminating the blind spots that arise between scheduled scans. We read this as an effort to embed AI‑driven security deeper into DevOps pipelines, especially for organizations already using Claude Enterprise. The automatic fix detection and reopening features could reduce the operational overhead for security teams, making the product more attractive to enterprises with large, fast‑moving codebases. The eight‑repository cap may limit adoption among very large firms, but the credit incentive could spur early experimentation that drives broader integration via API and MCP. Overall, the move hints at a growing convergence between AI models and security tooling, which could reshape how vulnerabilities are managed across both crypto and traditional software projects.
Sources used for this analysis
DeFi Intel wrote this analysis from the reporting above. Any figure or quotation can be checked against those pages.
Frequently asked questions
What is the main change introduced in Claude Security?
Claude Security now continuously monitors connected repositories and rescans automatically whenever a new commit is made, replacing the previous calendar‑based scanning schedule.
How does Claude Security handle vulnerability resolution?
Each identified vulnerability is tracked individually; when the tool detects the code change that fixes the issue, it automatically marks the finding as resolved, and it can reopen the issue if the same flaw reappears later.
What incentives does Anthropic offer for using Claude Security?
Eligible Claude Enterprise customers can receive up to $25,000 in monthly usage credits for two months, with the offer remaining open until December 31, 2026.
Direct answers
Short answers, written to be quoted.
- How does the new Claude Security scanning work?
- The new Claude Security scans continuously, running a deep baseline scan when a repository is first connected and then performing lighter scans focused on each new commit.
- What happens when Claude Security detects a fix for a vulnerability?
- When Claude Security detects the code change that resolves a vulnerability, it automatically marks that finding as fixed and can reopen it if the issue reappears.
DeFi Intel is an entity-graph aggregator: we curate, tag and link crypto news to a typed knowledge graph of protocols, tokens, people and incidents. We do not republish the full article body. Use the link above to read the original report at Crypto Briefing.
Entities in this story
Discussion
Comments are reviewed before publication.
Want the full article?
Continue reading on Crypto Briefing →
No comments yet.