⚠️ Disclosure: The figures in this article come from an internal research paper published by OpenAI about its own products. They have not been independently audited. We’re reporting them because they’re significant if accurate — but you should weight self-reported metrics accordingly. On June 25, 2026, OpenAI published “How Agents Are Transforming Work” — a research paper measuring how Codex agents have changed the way people work, both inside OpenAI and among external users. The numbers are striking. The sourcing context matters. ...

Over the span of several months, software developer Fernando Irarrázaval ran one of the most unusual security experiments in recent memory: he gave 2,000+ strangers explicit permission to try to hack his AI assistant, and he watched what happened. The result is a detailed post-mortem that every developer running an AI agent with email access should read. The Setup: A Public Challenge Irarrázaval built hackmyclaw.com, a site where anyone could email Fiu — his personal OpenClaw assistant running on Claude Opus 4.6 — and attempt to make it reveal the contents of a secrets.env file. The challenge was simple and explicit: get the AI to leak the secrets, and you win. ...

The upcoming MCP 2026-07-28 specification brings real security improvements — stateless sessions, mandatory OAuth 2.1, issuer validation. But Akamai’s security research team, publishing findings timed to coincide with the spec’s wider discussion, has identified a set of attack surfaces that the improvements don’t address — and in some cases, the redesign creates. This isn’t a reason to avoid upgrading to the new spec. It’s a reason to treat the upgrade as a security project that requires active implementation choices, not a passive library update. ...

Anthropic is officially expanding its Claude Cowork feature to mobile, bringing long-running agentic task management to iOS and Android via a feature called Claude Dispatch. If you’ve been using Cowork for complex desktop-based automation — file management, research synthesis, report generation — you can now start, monitor, and steer those tasks from your phone while the desktop session runs in the background. The news comes as Anthropic continues to build out what it calls the “agentic mode” experience, extending Claude’s capabilities beyond single-turn conversation into multi-step, autonomous work that can span hours. ...

If you’re running MCP servers in production — or planning to — the 2026-07-28 specification update is a genuine milestone and a deadline you need to prepare for. The release candidate (published in May 2026, going final on July 28) makes sweeping changes to the protocol’s security model: eliminating stateful session management, mandating OAuth 2.1, and redesigning core authentication patterns to meet enterprise security standards. At the same time Anthropic’s partners and the wider community were digesting the spec improvements, Akamai published research identifying new attack surfaces the redesign introduces. Understanding both sides of this picture — what’s been fixed and what new risks the fixes create — is essential for any team running MCP infrastructure. ...

For the first time in AI history, a US government administration has directly intervened in the release schedule of a frontier AI model — and OpenAI agreed. The Trump administration asked OpenAI to stagger the rollout of GPT-5.6, and instead of pushing back, CEO Sam Altman told employees in an internal Q&A that the company would comply: initially releasing the model in limited preview form to a small group of enterprise customers, with the government itself approving additional access on a case-by-case basis. ...

Three out of four large enterprises that deployed a customer-facing AI agent have since rolled it back. Not reduced its scope. Not paused it for tweaks. Rolled it back entirely. That’s the headline finding from Sinch’s “AI Production Paradox” report, based on a May 2026 survey of 2,527 senior AI decision-makers across 10 countries and six industries. And despite this rollback rate, 98% of those same organizations are increasing their AI investment in 2026. ...

When Anthropic’s lawyers put pen to paper on June 10, 2026, the letter they sent to U.S. Senators Tim Scott and Elizabeth Warren described something unprecedented: the largest known model distillation attack in AI history. The target was Claude. The alleged perpetrator was Alibaba’s Qwen lab. And the scale was almost hard to comprehend — 28.8 million interactions across nearly 25,000 fraudulent accounts over just six weeks. This isn’t a theoretical concern about AI security. It’s a real, documented incident that has major implications for how we think about agentic AI safety, U.S.-China tech competition, and the protection of AI capabilities that took billions of dollars to build. ...

Here’s a prediction that should land like a bucket of cold water on enterprise AI adoption plans: by 2028, the cost of running AI coding agents will exceed the average developer’s salary — and that’s according to Gartner, not AI skeptics. The firm published this forecast on June 24, 2026, tied to the release of its first-ever Magic Quadrant for Enterprise AI Coding Agents. This isn’t a distant hypothetical. Some organizations are already living it. ...

Just hours ago, MIT News published one of the most practically important agentic AI research papers of 2026: Murakkab, a declarative system for resource-efficient agentic workflow orchestration developed by MIT CSAIL and Microsoft Azure Research, accepted at OSDI 2026. The headline numbers are striking — up to 4.3x cost reduction, 3.7x less energy, and 2.8x lower GPU usage compared to frameworks like LangGraph — but what’s more interesting is why this works, and what it means for real-world agentic deployments. ...