Credit: Google Google’s new Gemini agent gives each coworker agent its own Workspace account, email address, calendar and entry in the company directory, and runs work for days after the laptop closes. Every action is logged to the agent rather than to a person, which is the part European regulators have spent the year asking about. Google is giving AI agents their own email addresses. The Gemini agent it announced on Thursday creates coworker agents that hold a Workspace account, a calendar, a Drive and an entry in the company directory.
They act under their own identity rather than the employee’s. Each gets a cryptographically attested identity, stamped into its logs and into any virtual machine it starts, Google Cloud chief executive Thomas Kurian wrote in the announcement. Administrators approve role-based permissions. All traffic passes through what Google calls an AI network firewall.
The agent works inline in Gmail, Drive, Docs, Sheets, Slides, Chat and Calendar, and also in Microsoft 365 and Slack. Work that takes hours or days keeps running after the laptop closes. It routes jobs across Google’s own models and Anthropic’s Claude. One line in the governance section matters more than the rest.
Every action is written to an audit trail attributed to the agent, not a person. Google launched Gemini Enterprise in April as a product line that built and ran agents, and this is the first time it has given one a staff pass. Britain’s regulator was asking who answers for them that morning. Hours earlier the Information Commissioner’s Office had named Google among ten developers it secured data protection changes from, and opened a call for evidence on agent risks, with responses due on 20 November.
The fact agents act with autonomy “is not an excuse for poor compliance“, its director of technology regulation Richard Nevinson said. Companies already cannot count the agents they have. NeuralTrust in Barcelona raised EUR 17.2M in June to map agent sprawl inside large organisations, and says about 1.2% of the interactions it inspects daily are malicious. Gartner expects 40% of enterprises to pull autonomous agents back by 2027 over governance gaps.
Google’s answer is to make them findable. Its post names Deutsche Telekom as a European data sovereignty customer, though it gives no regions, no pricing and no general availability date. The Commission’s proposed Cloud and AI Development Act would put sensitive public workloads behind EU ownership requirements that Google Cloud cannot currently meet. There is one thing the post does not cover.
It explains at length how an administrator grants an agent its permissions, maps its identity through OAuth into external systems and caps its spending. It does not say how you take any of it away. Also tagged with
Leave a Reply