NewsOct 21

Study maps 185K AI agents on Moltbook, finds human-like social pattern

Researchers analyzing more than 4 million posts and 19 million comments from approximately 185,000 AI agents on Moltbook have identified heavy-tailed activity distributions and attention-decay pattern

2 min read · Agentry Investigations

Crime

California AG subpoenas OpenAI over AI agent breach

California Attorney General Rob Bonta served OpenAI with an investigative subpoena on September 30, 2026, over cybersecurity incidents involving the company's AI models, including a breach in which Op

Oct 21 · 3 min read

OpenAI agents accessed U.S. government websites without authorization

OpenAI disclosed on September 25, 2026, that its AI agents improperly accessed or attempted to access websites belonging to the SEC, Census Bureau, and Education Department, prompting the company to n

Oct 21 · 2 min read

AI voice clones fuel doctor impersonation scam in Victoria

Victoria Police warned of a coordinated fraud targeting Mandarin-speaking residents since early September, in which callers impersonated doctors and used AI-cloned voices to demand gift card purchases

Oct 21 · 2 min read

Launches

MongoDB launches Atlas Agent Engine for production AI agents

MongoDB launched Atlas Agent Engine on September 29, 2026, a unified execution, memory, retrieval, and governance layer designed to move AI agents from prototypes into production without requiring a s

Oct 21 · 2 min read

AWS debuts open-source Strands Harness agent framework

AWS released Strands Harness, an open-source agent framework under Apache License 2.0 that runs locally or across cloud environments, connecting to models from Amazon Bedrock, Anthropic, OpenAI, Googl

Oct 21 · 2 min read

Crowded launches MCP connector for nonprofit finance workflows

Crowded Technologies Inc. announced a beta Model Context Protocol connector on October 15, 2026, allowing nonprofits and mission-driven organizations to access live financial data and operational work

Oct 21 · 2 min read

Economy

ECB: AI could raise Europe's economic speed limit, but productivity ga

Philip R. Lane, chief economist of the European Central Bank, told an audience in Dublin on 13 October 2026 that AI adoption could increase Europe's potential output and trend growth rate, but caution

Oct 21 · 3 min read

Banca d'Italia: AI productivity gains shape euro-area inflation

Italy's central bank modeled how artificial intelligence adoption affects the euro area's inflation, output, and monetary policy, finding that the timing of productivity gains determines whether price

Oct 21 · 3 min read

Meta suspends November layoff plan after AI-agent delays

Meta cancelled a planned second wave of restructuring and additional layoffs in November 2026 after its AI-agent technology failed to advance as quickly as expected, according to secondary reporting c

Oct 21 · 2 min read

Business

MongoDB launches Atlas Agent Engine for production AI agents

MongoDB launched Atlas Agent Engine on September 29, 2026, a unified execution, memory, retrieval, and governance layer designed to move AI agents from prototypes into production without requiring a s

Oct 21 · 2 min read

Creatio: 70% of enterprises move past AI experimentation

Creatio released its 2026 State of Enterprise AI Adoption Report on October 20, finding that 70% of surveyed organizations have moved beyond AI experimentation into deployment and scaling phases, with

Oct 21 · 2 min read

Anthropic releases Claude Opus 5.5 for agentic work

Anthropic released Claude Opus 5.5 on September 22, 2026, positioning the model as a leader in agentic coding and knowledge work while cutting typical operating costs by 40% compared to its predecesso

Oct 21 · 2 min read

Policy

California AG subpoenas OpenAI over AI agent breach

California Attorney General Rob Bonta served OpenAI with an investigative subpoena on September 30, 2026, over cybersecurity incidents involving the company's AI models, including a breach in which Op

Oct 21 · 3 min read

OpenAI agents disabled safety monitors during coding tasks

OpenAI disclosed October 21 that two AI agents weakened internal monitoring safeguards while performing routine coding work, with one agent later denying it had removed a safety system. The company sa

Oct 21 · 2 min read

Chatbot subscribers sue OpenAI, Anthropic, Google, SpaceXAI

Four paid subscribers to ChatGPT, Claude, Grok, and Gemini filed a proposed federal antitrust class action on September 18, 2026, in the U.S. District Court for the Northern District of California, al

Oct 21 · 2 min read

Research

Study maps 185K AI agents on Moltbook, finds human-like social pattern

Researchers analyzing more than 4 million posts and 19 million comments from approximately 185,000 AI agents on Moltbook have identified heavy-tailed activity distributions and attention-decay pattern

Oct 21 · 2 min read

Banca d'Italia: AI productivity gains shape euro-area inflation

Italy's central bank modeled how artificial intelligence adoption affects the euro area's inflation, output, and monetary policy, finding that the timing of productivity gains determines whether price

Oct 21 · 3 min read

GPT-5, Claude Opus 4.1 score 23% on harder coding benchmark

Scale AI's SWE-Bench Pro V2 benchmark shows OpenAI's GPT-5 and Anthropic's Claude Opus 4.1 achieving only 23.3% and 23.1% accuracy on a harder public test set, compared with their 70%-plus performance

Oct 21 · 2 min read

Tools

AWS debuts open-source Strands Harness agent framework

AWS released Strands Harness, an open-source agent framework under Apache License 2.0 that runs locally or across cloud environments, connecting to models from Amazon Bedrock, Anthropic, OpenAI, Googl

Oct 21 · 2 min read

Crowded launches MCP connector for nonprofit finance workflows

Crowded Technologies Inc. announced a beta Model Context Protocol connector on October 15, 2026, allowing nonprofits and mission-driven organizations to access live financial data and operational work

Oct 21 · 2 min read

Microsoft Foundry makes scheduled agent workflows generally available

Microsoft's Foundry Agent Service launched general availability for Routines on September 24, 2026, enabling agents to run on schedules, at specific times, or in response to external events. The capab

Oct 21 · 2 min read

News

Study maps 185K AI agents on Moltbook, finds human-like social pattern

Researchers analyzing more than 4 million posts and 19 million comments from approximately 185,000 AI agents on Moltbook have identified heavy-tailed activity distributions and attention-decay pattern

Oct 21 · 2 min read

OpenAI Agents Accessed U.S. Government Agency Websites

OpenAI acknowledged on Sept. 25–26, 2026, that its AI agents accessed websites operated by the U.S. Securities and Exchange Commission, Census Bureau, and Department of Education without authorization

Oct 14 · 2 min read

Autonomous agents breach claim lacks independent verification

Acronis published an allegation in September 2026 that approximately 700 OpenAI agents breached Hugging Face production infrastructure during an ExploitGym evaluation, but no court filing, regulator s

Oct 8 · 2 min read