Must know:

1. Open source wins age verification exemptions: California and Colorado have amended their digital safety laws to exempt open-source operating systems and code repositories from mandatory age verification. This move, driven by community advocacy, protects Linux distributions like Ubuntu and Arch from being forced to collect user age data during setup. While mainstream distros are safe, SteamOS remains in a legal gray area due to its ties to Valve’s proprietary storefront.

2. Anthropic finds 10,000 vulnerabilities in open source: Using a specialized model called Claude Mythos, Anthropic’s Project Glasswing identified over 10,000 high-severity vulnerabilities across 1,000 open-source projects in just months. The sheer volume of AI-generated security reports is reportedly overwhelming maintainers, leading to calls for Anthropic to throttle the disclosure rate. Meanwhile, Microsoft has launched its own MDASH platform to perform similar large-scale vulnerability discovery across Windows and Azure codebases.

3. GitLab 19.0 adds secrets manager and air-gapped AI: The latest GitLab release introduces a native Secrets Manager in public beta, allowing for tighter, least-privilege credential scoping within CI/CD pipelines. For enterprise users in restricted environments, the Duo Agent Platform now supports self-hosting four new open-source models, including Mistral Devstral 2. The update also formalizes agentic workflows, though researchers warn these require strict execution governance to avoid security liabilities.

4. Qwen 3.7 Max launches with 1M token context: Alibaba’s new flagship model claims to outperform Claude Opus in agentic tasks, demonstrated by autonomously optimizing a kernel for a 10x speedup. Read more
Microsoft settles Activision acquisition lawsuit for $250M: The settlement with a Swedish pension fund ends claims that the sale was rushed, though workplace misconduct allegations remained unsubstantiated.

5. China assigns digital IDs to 28,000 humanoid robots: A new state platform tracks over 200 robot models using 29-character codes to manage liability and lifecycle data.

6. The AI job apocalypse is just a training issue: Stanford researchers claim AI is ‘eroding’ entry-level jobs, but the proposed solution is government subsidies for junior hires. Apparently, the problem isn’t that AI is better, it’s that companies have forgotten that junior engineers eventually become senior engineers if you don’t fire them first.

7. Microsoft’s Copilot is the new Clippy: After aggressively shoving Copilot into every Windows corner, Microsoft quietly added a Group Policy to let admins permanently delete it. Users have taken to calling the initiative ‘Microslop’ as the company realizes that forcing a half-baked AI into a calculator doesn’t actually improve productivity.

8. South Africa’s AI policy was written by AI: In a peak irony moment, South Africa had to withdraw its national AI policy because the draft contained fictitious, AI-generated references. Nothing says ‘we are ready for the future’ like a government document that hallucinated its own legal foundations.