Telecommunications / Fixed and mobile telecom
Comcast: AI security vulnerability assessment and remediation across 258 business-critical systems
- Company
- Comcast
- Country
- United States
- Adoption stage
- Limited operation
- Source published
- Date basis
- The date the source was published. It can differ from the date adoption started.
- How the source was checked
- Read the full source text
The work problem
According to an article Anthropic published on October 6, 2026, existing security tools can identify single vulnerabilities in isolation, but exploit chains, in which an attacker links several individually minor weaknesses into a path to something serious, are much harder to trace. Comcast's Chief Information Security Officer said the volume of findings produced with AI is enormous and that validating them has become the new bottleneck.
Technology and data
Through Project Glasswing, Anthropic's limited access program, Comcast used Claude Mythos Preview to assess 258 business-critical systems and approximately 170 million lines of code. Instead of evaluating files or components one at a time, the model keeps the context of code, configuration and application behavior together to find security gaps that emerge across multiple components. Comcast engineers validated the model's finding against the running application, traced the root cause, implemented remediation and verified that the issue had not been exploited before it was fixed.
Results
In this assessment, Claude Mythos Preview identified a critical authentication vulnerability in a public-facing platform that could have allowed an attacker to bypass authentication controls, and Comcast remediated it before any evidence of exploitation was observed. The vulnerability arose from the interaction of multiple systems and assumptions that appeared correct when reviewed independently. Comcast's Chief Information Security Officer said early access accelerated the company's understanding of the security capabilities of frontier AI models by about a year.
Limits and open questions
The length of the assessment, the total number of vulnerabilities found and the share of incorrect findings are not in the source.
Sources
Compiled from public sources. These are not results from ATF Works customers.