🐘 OpenAI's models find PHP flaws and the US bans DeepSeek

⚡ aiBytes_ #15

Hello, builders 👋

Welcome back to your weekly dose of AI Bytes. I did not expect PHP to be the biggest AI story of the week. But OpenAI's models found critical flaws in the language that runs an estimated 70% of websites, and a coalition of maintainers spent two months fixing them before anyone else knew 🐘

AI is becoming the security team: the same week, Anthropic launched OSS Scanner, which gives critical open-source projects free vulnerability scans from its strongest models, with a proof of concept for each finding and a patch where it can. For years the worry was that AI would help attackers first. This week, the defenders got there first.

In this issue

🌍	AI News	8 stories
📣	Official Announcements	Anthropic
0x02	Viral on X	5 posts
0x03	New AI Products	4 launches
0x04	Trending Github Projects	5 repos
0x05	HN Deep Stories	5 threads

🐘 OpenAI's models found critical security flaws in PHP - The language behind an estimated 70% of websites. Every PHP site can now safely upgrade. Story of the week (OpenAI)

🚫 DeepSeek banned from US federal civilian agencies - A new DHS order pulls DeepSeek out of government agency IT systems. (Reuters)

⚡ DeepInfer - DeepSeek's own high-performance inference framework for its models, on GitHub. (HackerNews)

🔎 DeepSeek search and fetch - Web search and page fetch in the DeepSeek API, at $0.75 per 1,000 successful requests. (HackerNews)

💰 OpenAI is working on a $30B raise at a $1.25T valuation - It would be the most highly valued private company ever, ahead of an IPO. (TechCrunch)

🧭 OpenAI makes its Jev competitor widely available - OpenAI's decision model opens to everyone as the race with Jev heats up. (TechCrunch)

🔐 1Password lets AI agents log in for you - Agents can sign in to your accounts without ever seeing the password. (TechCrunch)

🎙️ Voxtral 3 - Mistral releases the third generation of its Voxtral speech models. (HackerNews)

This week's launches, in the companies' own words.

Anthropic @ClaudeDevs · MON 19 We launched OSS Scanner, which gives critical open source projects free, periodic vulnerability scans from our strongest models. Each finding includes a proof-of-concept, and a patch where possible. https://red.anthropic.com/oss-scanner Post on X ↗ · red.anthropic.com ↗

Anthropic @ClaudeDevs · THU 22 We're donating Agent Skills to the Agentic AI Foundation (@AgenticAIFdn), the same foundation we donated MCP to. Since December 2025, 70+ agent platforms have integrated the standard, including ChatGPT, Cursor, GitHub Copilot, and VS Code. Here's what that means for you: Post on X ↗

Anthropic @AnthropicAI · SUN 18 We’re committing a portion of Anthropic’s shares to the Anthropic Public Benefit Fund, a new wholly owned entity dedicated entirely to the public benefit. It will start off with 5% of our shares. For every 50% increase in the company’s value over a set baseline, we can contribute up to an extra 2%, subject to our board’s discretion. … Post on X ↗