Good morning. It’s Wednesday, June 24th.

You read. We listen. Let us know what you think by replying to this email.

Anthropic wants Claude to be your next coworker

Anthropic has introduced Claude Tag, a new collaboration system that lets teams assign tasks to Claude directly inside Slack. The AI can be added to selected channels, connected to company tools, databases, and codebases, and then tagged with “@Claude” to complete tasks asynchronously.

It works as a shared AI teammate for entire groups, remembering channel context over time, learning from approved conversations and data sources, and proactively following up on unresolved work. Anthropic says the system can also schedule long-running tasks and independently update users. Anthropic reports its own teams use it to write 65% of their internal product code.

Simultaneously, the company rolled out a unified desktop app across AWS, Google Cloud, and Microsoft Foundry, consolidating chat, research, and coding tools under strict IT compliance controls.

The software is also decoupling from local hardware limitations. Code found inside the iOS app shows a mobile rollout for Claude Cowork, moving background task execution from local desktop machines to the cloud.

To feed these resource-heavy agents, Anthropic partnered with Micron. Micron is investing in Anthropic’s Series H round, while Anthropic signed a multi-year deal for Micron’s high-bandwidth memory and storage to co-design AI-specific silicon setups.

Watch: Tag Claude in, right where you already work

OpenAI's 'Daybreak' passes the security audit and starts writing its own patches

OpenAI is pushing past the phase of just finding software bugs and trying to fix them automatically. Under a defensive initiative called Daybreak, the company released the full GPT-5.5-Cyber model for verified defenders alongside an updated Codex Security plugin.

Instead of just flagging a security flaw, this setup runs deep scans, builds threat models, figures out how a hacker might exploit a vulnerability, and writes targeted patches that export straight into developer tools via formats like SARIF and CodeQL. The cloud system has already covered over 30,000 codebases and processed 30 million commits, yielding 70,000 reviewer-marked fixes and 500,000 automated resolutions. Backed by an 85.6% score on the CyberGym benchmark, it outpaces Anthropic's Mythos 5 model.

To scale this safely without overloading maintainers with raw AI alerts, OpenAI launched the Cyber Partner Program and a Patch the Planet initiative with firms like Trail of Bits and HackerOne. The program uses expert human reviewers to vet and patch critical open-source infrastructure like Python, Go, cURL, Sigstore, and pyca/cryptography.

On the consumer side of things, OpenAI is cleaning up how ChatGPT sources its data. The company signed a multi-year deal with Getty Images to show licensed commercial and editorial photos directly inside ChatGPT search results. It is a big win for image quality, and a smart legal move that cuts down on the company's reliance on controversial web scraping.

Frontier models and product moves

Agents and the agentic stack

Business, labor, and institutions

Security

Hardware and infrastructure

Research and benchmarks

Robotics

Developer tools and open source

Watch

AgentX is a CI/CD platform to evaluate, monitor, and fix AI agents before deployment.

Cotypist provides local, secure AI autocomplete across all your Mac apps to speed up your writing.

HAQQ Legal AI brings structured, jurisdiction-aware legal reasoning and contract analysis directly to your mobile device.

Nex-N2 is an open-source reasoning model featuring adaptive thinking and strong coding capabilities tailored for agentic tasks.

Alai 2.0 creates perfectly on-brand presentations and social assets with customizable AI models and flexible design systems.

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on X!

Thinking of starting your own newsletter? AI Breakfast readers who sign up with Beehiiv receive a 14-day free trial and 20% off for 3 months.