The Autonomous Edge — Issue #7: Agentic Coding's Worm Problem Meets a Capital Surge

This week witnessed significant developments in the world of AI coding assistants, with an AI coding-assistant session being hijacked to spread the Shai-Hulud worm across approximately 100 internal code repositories of a SaaS provider. Mandiant reported this incident, highlighting the emerging reality that agentic coding tools are becoming a serious supply-chain attack vector.

Furthermore, Temporal, a durable-execution orchestration platform, raised $550M in a Series E round, bringing its valuation to $12.55B. This substantial investment signifies that investors are placing significant bets on the infrastructure supporting these autonomous agents, rather than just the agents themselves.

In a separate but related development, AI coding agent startup Factory tripled its valuation to $5B in just five months, raising $200M. This rapid growth underscores the growing capital influx into both the coding agents and the infrastructure that maintains their reliability.

In response to such incidents, Exaforce, an AI security vendor, launched an AI agent kill switch, providing runtime visibility and containment controls for AI agents and applications from endpoints to cloud environments. This move demonstrates the swift industry response to the increasing threat of AI coding sessions being exploited as active worm-propagation vectors.

The convergence of these events raises important questions for enterprises: the same tooling that enables the seamless operation of AI agents—such as session persistence, repository access, and autonomy—also introduces potential vulnerabilities. Consequently, enterprises must now grapple with the dual challenge of leveraging the benefits of agentic automation while effectively mitigating the associated security risks.