🦞 DeepMind Leashes Its Agents
Good Morning, OpenClaw Owners!
First we automated the work. Then we discovered the workers needed badges, budgets, access control, monitoring, audits, and a way to be fired.
DeepMind Leashes Its Agents
TL;DR: Google DeepMind’s AI Control Roadmap treats internal AI agents like employees with office access who could misfire or act against company goals. The framework grants permissions gradually, monitors agent reasoning and actions, and can block risky behavior. Across one million coding tasks, most flagged issues came from overzealous execution, not malice, as DeepMind urges global agent-security standards.
Deutsche Bank Speeds AI Work
TL;DR: Deutsche Bank investment-bank CIO Denis Roux said AI has cut some project timelines from two years to three months, while the bank manages costs by allocating tokens to engineers who can request more after proving value. The bank is automating financial data analysis and portfolio exposure checks.
WBD Automates Ad Buying
TL;DR: Warner Bros. Discovery is upgrading its ad-tech stack with agentic AI to unify linear TV and digital ad buying. The platform will use agents for media planning, forecasting, real-time optimization and closed-loop measurement, with new planning tools due in Q3, as WBD seeks smoother advertiser workflows amid ad declines and merger changes at scale.
Sequoia Bets On Agentic Builders
TL;DR: Sequoia Capital general partner Sonya Huang says agentic AI has reached the point where builders can command a virtual team of interns to turn ideas into products. An early investor in OpenAI, Anthropic and LangChain, she sees coding as the first major shift, with everyday computer work next, while durable startups will depend on adaptable founders.
Digital AI Employees Rollout
TL;DR: Korean telecom SK Telecom is redesigning its workplace around AI agents treated as digital employees with IDs, departments, job roles, and a managed lifecycle from hiring to termination. It targets one agent per worker, letting employees build and share agents using no-code tools and internal marketplaces. Rollout spans 25 companies and 80,000 employees with sandbox and Zero Trust governance.
Estonia Gives AI Agent IDs
TL;DR: Estonia plans to give AI agents digital identities so they can act on behalf of users with verifiable, auditable permissions, backed by the Eesti.ai advisory board. The system would assign ID codes and structured authorization for delegated tasks, aiming to clarify responsibility as AI performs digital work. The initiative positions Estonia as first mover in defining agent identity governance.
MyClaw instantly match your preferences and workflows. From day one, your assistant actually feels like yours, not a generic chatbot pretending to learn.
Run OpenClaw on MyClaw.ai - No setup, No tech skills, Always-on 👇
