BT has become the first UK company to join Anthropic's Project Glasswing, an initiative that deploys advanced artificial intelligence to defend against rapidly evolving cyberattacks. The telecom giant will use Anthropic's Claude Mythos Preview AI model to analyze network traffic and detect threats in real time, marking a significant shift in how critical infrastructure operators approach cybersecurity.

Project Glasswing, launched by Anthropic earlier this year, focuses on integrating large language models into security operations. The partnership with BT signals growing adoption of generative AI for defensive cyber measures, an area that has traditionally relied on rule-based detection systems and signature matching.

The AI Security Shift

BT's adoption of Claude Mythos Preview reflects a broader trend in the cybersecurity industry. Traditional security tools often fail against novel attack patterns. AI models trained on vast datasets can identify anomalies that static rules miss. The BT deployment will focus on protecting the company's sprawling network, which serves millions of residential and business customers across the UK.

Anthropic's Claude model brings natural language understanding and reasoning capabilities to threat detection. This allows security teams to query the system about suspicious activity and receive contextual explanations rather than simple alerts. The model can also adapt to new attack vectors without requiring frequent manual updates.

Why This Matters

Telecommunications networks are prime targets for state-sponsored hackers and ransomware groups. A successful breach could disrupt internet access, emergency services, or financial transactions. By embedding AI directly into its defense infrastructure, BT aims to reduce detection times and improve response accuracy. For UK consumers and businesses, this could mean fewer service outages and stronger protection against data theft. The partnership also puts pressure on other telecom operators to invest in similar AI-driven security solutions.

The move comes amid a surge in cyberattacks on critical national infrastructure. In 2024, the UK’s National Cyber Security Centre reported a 30% increase in incidents targeting telecom providers. AI tools like Claude Mythos Preview offer a way to scale defense without proportionally increasing human analyst headcount.

Anthropic’s Expanding Role

Anthropic, a San Francisco-based AI safety startup, has positioned itself as a leader in responsible AI deployment. Project Glasswing is part of its effort to showcase practical applications of Claude beyond general-purpose chatbots. The partnership with BT gives Anthropic a foothold in the European telecom sector, a market worth billions in cybersecurity spending annually.

The collaboration also highlights a growing trend: large telecom operators turning to external AI providers rather than building their own models. This approach offers faster deployment and access to cutting-edge research but raises questions about vendor dependency and data sovereignty. BT has stated that all network data processed by Claude Mythos Preview will remain within the UK and comply with local regulations.

What Comes Next

BT plans to expand the use of Claude Mythos Preview across additional security functions over the next year. The company is also exploring how the model can assist in compliance monitoring and incident reporting. If successful, the partnership could serve as a template for other UK infrastructure providers looking to modernize their cyber defenses.

The broader cybersecurity industry will watch closely. Generative AI models have proven effective at analyzing text and code, but their application to real-time network defense is still unproven at scale. BT’s deployment will provide valuable data on reliability, latency, and false positive rates. The outcome could influence how regulators view AI in critical systems.