- An Anthropic exec believes that fully autonomous “virtual employees”—AIs with their own logins, memories, and job roles—could hit corporate networks within 12 months. He framed the shift as a step beyond today’s task-based agents.
- He warned most companies lack controls for non-human identities, raising hard questions about permissions, monitoring, and liability. Anthropic is investing in tools to fingerprint AI account behavior and urges enterprises to do the same.
- Businesses should treat any AI agent with credentials like a high-risk hire and impose zero-trust, kill-switch policies before deployment.