Devlery

Blog

Notes and analysis on AI development.

No AI model license, but a 30-day security testing window

No AI model license, but a 30-day security testing window

Trump’s new AI cybersecurity order avoids mandatory model licensing while creating a 30-day clearinghouse and a voluntary frontier model cyber evaluation path.

Microsoft Aion 1.0 Plan Brings a 14B Local Agent to Windows

Microsoft Aion 1.0 Plan Brings a 14B Local Agent to Windows

Microsoft tied Aion 1.0 Plan, Edge on-device AI APIs, Windows AI APIs, and MXC into a local agent path for Windows builders.

Sema4 Agent Builder turns SOPs into back-office agent runbooks

Sema4 Agent Builder turns SOPs into back-office agent runbooks

Sema4.ai introduced Agent Builder, MCP Access Gallery, Business Context Layer, verified queries, and audit logs for enterprise back-office agents.

Microsoft ASSERT and ACS split agent policy into tests and runtime blocks

Microsoft ASSERT and ACS split agent policy into tests and runtime blocks

Microsoft introduced ASSERT and ACS for agent governance, turning natural-language policy into evaluation artifacts and runtime enforcement points.

Anthropic turns Claude consulting partners into a public scorecard

Anthropic turns Claude consulting partners into a public scorecard

Anthropic added a Services Track and Partner Hub to the Claude Partner Network, ranking consultants by certification, production customers, and public references.

Google AX preview turns interrupted agents into resumable runtime work

Google AX preview turns interrupted agents into resumable runtime work

Google AX is an open source runtime for long-running agents, covering resumption, isolation, auditability, and Kubernetes deployment.

Meta expands Business Agent, putting WhatsApp partners on paid-platform terms

Meta expands Business Agent, putting WhatsApp partners on paid-platform terms

Meta Business Agent now handles customer conversations across WhatsApp, Messenger, and Instagram, changing the competitive terms for BSPs and AI commerce tools.

Microsoft Foundry adds an ROI loop for production AI agents

Microsoft Foundry adds an ROI loop for production AI agents

Microsoft Foundry now ties agent traces, rubric evaluation, optimization, and ROI dashboards into one Agent DevOps loop.

Work IQ APIs go GA on June 16 with a Microsoft 365 agent pricing model

Work IQ APIs go GA on June 16 with a Microsoft 365 agent pricing model

Microsoft Work IQ APIs reach GA on June 16. Here is what M365 context, MCP/A2A, delegated auth, and Copilot Credits mean for agent builders.

GPT-Rosalind hits 63.2% as Codex adds life-science plugins

GPT-Rosalind hits 63.2% as Codex adds life-science plugins

OpenAI updated GPT-Rosalind with LifeSciBench, MedChemBench, GeneBench, LabWorkBench scores and Codex life-science plugins.

VS Code Copilot Sessions Add Remote Agents and Terminal Safety

VS Code Copilot Sessions Add Remote Agents and Terminal Safety

GitHub Copilot’s May VS Code updates bundle Agents Window, remote sessions, BYOK controls, and terminal risk checks.

Anthropic maps 832 blocked accounts to agent attacks MITRE does not yet name

Anthropic maps 832 blocked accounts to agent attacks MITRE does not yet name

Anthropic Red Team mapped 832 blocked accounts and 13,873 observations to MITRE ATT&CK, exposing gaps around AI agent orchestration.