October 7, 2026
Anthropic is expanding its cybersecurity offering with a new version of its Cyber Verification Program, giving qualifying security professionals access to more powerful Claude models and fewer restrictions on advanced cybersecurity capabilities.
The move reflects a growing reality in the AI industry.
The same technology that can help developers find vulnerabilities can also be used by attackers.
AI companies are therefore trying to give legitimate security researchers powerful tools while preventing those capabilities from being abused.
Claude becomes a more serious cybersecurity assistant
Anthropic says the expanded program provides three levels of access for qualifying security professionals.
The program includes access to models such as Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1, along with future models.
For security teams, advanced AI can help with tasks that would traditionally require significant manual work.
That includes analysing code, identifying vulnerabilities and examining large quantities of technical information.
AI can now investigate software at a different scale
One reason this matters is the sheer volume of modern software.
Large companies can have millions of lines of code spread across thousands of applications and dependencies.
Human security teams cannot inspect everything manually.
AI can potentially analyse enormous amounts of code much faster and identify suspicious patterns that deserve human attention.
That does not mean AI can replace security researchers.
Instead, it changes what researchers can realistically investigate.
But the same capabilities can be dangerous
Cybersecurity is one of the most sensitive areas of AI development.
A powerful model can help a security researcher understand a vulnerability.
The same model could potentially help someone exploit it.
That creates a difficult balancing act for AI companies.
Anthropic’s solution is to provide advanced capabilities through a controlled access program rather than simply making every capability available to everyone.
AI security is becoming an arms race
Anthropic is not alone.
OpenAI, Google, Microsoft and other technology companies are also investing heavily in AI-powered cybersecurity.
The competition is increasingly happening on both sides.
Defenders want AI to find vulnerabilities faster.
Attackers want AI to discover weaknesses faster.
That means cybersecurity may become one of the first fields where AI agents routinely work alongside professionals in real time.
What this means for ordinary users
Consumers may not interact directly with Anthropic’s cybersecurity tools.
But the technology could still affect them.
If AI makes vulnerability detection faster, companies may be able to discover and fix security problems before criminals exploit them.
At the same time, faster AI-assisted attacks could make cybersecurity more difficult.
The result is a new security environment in which both defenders and attackers have increasingly capable AI tools.
The cybersecurity battle is no longer simply between humans and software. It is increasingly becoming a battle between AI systems working for both sides.
Source: Anthropic – Project Glasswing and Cyber Verification Program
Laat een reactie achter