☕️ Microsoft and Google won't cut ties with Anthropic

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

🤝 Microsoft and Google won't cut ties with Anthropic

❌ Oracle and OpenAI end plans to expand flagship data center

👀 Trump unveils new US cybersecurity strategy

✍️ Grammarly accused of using identities without permission

🔒 OpenAI launches Codex Security

🕵️ FBI investigates hack of its wiretap system

Plus: 🎁 16 other news you might like, and 🧰 5 tools.

🤝 Microsoft and Google won't cut ties with Anthropic LINK

Microsoft and Google said their customers can still access Anthropic's AI tools, including chatbot Claude, through their platforms despite the Pentagon labeling the company "a supply chain risk."

The Pentagon blacklisted Anthropic after CEO Dario Amodei refused to give the US military "unrestricted access," specifically objecting to mass domestic surveillance and fully autonomous weapons uses.

Both companies said they can keep working with Anthropic on non-defense-related projects, while nearly 500 Google employees and 80 OpenAI staffers signed an open letter supporting Anthropic.

❌ Oracle and OpenAI end plans to expand flagship data center LINK

Oracle and OpenAI have dropped plans to expand a major AI data center in Abilene, Texas, after negotiations stalled over financing issues and shifts in OpenAI's needs.

Meta is now considering leasing the planned expansion site from developer Crusoe, with Nvidia helping facilitate those talks and paying a $150 million deposit to secure its chips would fill the facility.

The broader deal between Oracle and OpenAI for 4.5 gigawatts of data center capacity remains on track, and the companies have announced projects in other locations, including one near Detroit.

👀 Trump unveils new US cybersecurity strategy LINK

The White House released President Trump's seven-page cybersecurity strategy, which breaks from past approaches by putting offensive cyber operations — not just deterrence — at the center of US policy.

The strategy is built on six pillars, including disrupting adversaries before they attack, cutting back cyber regulations, modernizing federal networks with AI and zero-trust architecture, and securing critical infrastructure.

For the first time, a national cybersecurity strategy references cryptocurrencies and blockchain, while critics warn that pushing offensive operations and deregulation could expose critical systems and invite retaliation.

✍️ Grammarly accused of using identities without permission LINK

Grammarly's "expert review" feature generates AI writing advice labeled as "inspired by" named experts, including deceased scholars and working journalists, without asking any of those people for permission.

The Verge found its own editors — Nilay Patel, David Pierce, Sean Hollister, and Tom Warren — listed alongside dozens of tech journalists, with some profiles containing outdated job titles and other inaccuracies.

The feature's linked sources often led to spammy website copies or completely unrelated pages, suggesting the AI's advice attributed to one person may actually be based on someone else's work.

🔒 OpenAI launches Codex Security LINK

OpenAI has launched Codex Security, a new tool inside its Codex programming assistant that helps developers find and fix code vulnerabilities by scanning their repositories and ranking flaws by severity.

The tool copies a code repository into an isolated container, builds a threat model describing how the program works, then tests discovered flaws in a sandbox to filter out false positives.

Codex Security started as an internal OpenAI tool called Aardvark and is now available as a research preview for ChatGPT Enterprise, Business, and Edu tiers, plus free for open-source project maintainers.

🕵️ FBI investigates hack of its wiretap system LINK

The FBI is investigating a breach of its internal systems used to manage wiretaps and surveillance warrants, with suspected Chinese hackers potentially behind the intrusion into this sensitive law enforcement infrastructure.

Analysts have raised the possibility that the activity could be linked to the Salt Typhoon operation attributed to Chinese intelligence services, which previously targeted U.S. telecommunications and national security networks.

Officials say the issue has been contained but have released few technical details about how the suspicious activity occurred or whether any court records or case data was accessed.

Other news you might like

• Anthropic launches Claude Marketplace, giving enterprises access to Claude-powered tools from Replit, GitLab, Harvey and moreLINK

• Samsung teases what to expect from its upcoming smart glasses later this yearLINK

• Nintendo is suing the US government over Trump's tariffsLINK

• The MacBook Neo may cannibalize more iPads than MacBook AirsLINK

• Anthropic’s Claude found 22 vulnerabilities in Firefox over two weeksLINK

• Google's new command line tool can plug OpenClaw into your Workspace dataLINK

• Musk fails to block California data disclosure law he fears will ruin xAILINK

• Satellite firm pauses imagery after revealing Iran's attacks on US basesLINK

• Video AI models hit a reasoning ceiling that more training data alone won't fix, researchers sayLINK

• Anthropic's Claude Code subscription may consume up to $5,000 in compute per month while charging the user just $200LINK

• Polymarket and Kalshi’s CEOs Hate Each OtherLINK

• Engineer receives $30,000 for exposing a vulnerability affecting 7,000 robot vacuum cleaners — tinkerer just wanted to drive his robot vacuum with a PS5 controllerLINK

• X is testing a new ad format that connects posts with productsLINK

• Alphabet’s massive pay bet: Sundar Pichai could earn up to $692 millionLINK

🧰 Trending tools

TestSprite 2.1: automates frontend and backend testing workflows using natural language, handling test planning, code generation, execution, and debugging to reduce manual testing effort.LINK

Copperlane: an AI-powered loan origination platform that automates rate pricing, borrower guidance, and document verification to reduce processing time from hours to seconds.LINK

Codex Security: a platform for building and scaling AI-powered products using industry-leading models and integrated development tools for production deployments.LINK

FasterGH: a GitHub mirror with Convex-powered caching that delivers faster page loads by maintaining a real-time synchronized read model of your repositories.LINK

21st Agents SDK: a React-based development kit providing pre-built UI components, templates, and infrastructure for building AI agents with integrated tools, memory, and observability features.LINK

Want to get the latest news differently? Find us on:

See you tomorrow for a new dose of ☕️ Techpresso!