AI Coworkers Break Current Security Models

The rapid evolution of AI-driven work is introducing persistent AI coworkers, which require a fundamental shift in security models. Traditional access models, designed for task-scoped agents, are inadequate for these new digital colleagues. Persistent AI coworkers need independent access provisioning and lifecycle management, as they accumulate access over time, leading to potential security risks like privilege creep.

Current AI platforms lack proper identity management for agents, often relying on human credentials, which complicates audit trails and security. Platform vendors are beginning to adapt, offering solutions for agent-specific identities. Organizations must proactively manage AI coworkers by assigning them unique identities, defining clear ownership, and setting expiration conditions to mitigate risks.