Telecoms operators and vendors should prepare for the impact that LLMs will have on cyber security

On 7 April 2026, Anthropic announced the launch of its new large language model (LLM), Claude Mythos Preview, which Anthropic claims is unparalleled in its ability to find and exploit software vulnerabilities. The company and the media claim that this represents a major leap forward in cyber‑security capabilities, but these assertions appear somewhat exaggerated. Nonetheless, LLMs are gradually becoming increasingly effective at vulnerability detection and exploitation; this will have implications for telecoms operators and vendors.