AI Detente, Copilot Limits & Project Glasswing | Techstrong Gang
The Microsoft–OpenAI partnership just got rewritten. GitHub paused new Copilot sign-ups. Microsoft embedded Anthropic’s Claude into its security pipeline. All in 8 days.
On today’s Techstrong Gang, host Jon Swartz and the panel unpack what actually happened — and what it means for anyone shipping software in 2026.
AI Detente: OpenAI and Microsoft Rewrite the Deal
OpenAI just won multi-cloud freedom, with deals on the table for Amazon Web Services and Google Cloud, while Microsoft locked in long-term IP and revenue rights. The Azure exclusivity that defined the last three years of generative AI is officially over.
OpenAI is no longer Azure-only. Workloads can now spread across AWS and Google Cloud. Microsoft kept the IP rights and revenue share that matter long-term, keeping Redmond in the loop on every model generation. The exclusivity arrangement that powered Copilot, Azure OpenAI, and the Bing AI relaunch has been retired.
Copilot Hits the Brakes: GitHub Pauses New Sign-Ups
On April 20, GitHub paused new sign-ups for Copilot Pro, Pro+, and Student plans and tightened weekly usage caps for existing users. The reason: agentic coding sessions are burning more than 500,000 tokens per session versus roughly 100 for old-school autocomplete. The flat-rate twenty-dollar-a-month pricing model just broke under the load.
This isn’t just a GitHub problem. Cursor, Windsurf, Replit Agent, and every other AI coding tool will face the same math in 2026. Agentic coding burns 5,000 times more tokens than autocomplete, and flat-rate pricing can’t survive that reality.
Project Glasswing: Microsoft Embeds Claude Mythos in Its SDL
Microsoft is embedding Anthropic’s Claude Mythos Preview directly into its Security Development Lifecycle — the same lifecycle that gates Windows, Office, and Azure code. During preview, Mythos autonomously discovered a 17-year-old remote-code-execution flaw in FreeBSD’s NFS implementation, now tracked as CVE-2026-4747.
Anthropic has privately warned U.S. officials that the same capability in attacker hands makes large-scale cyberattacks “significantly more likely” in 2026. Defense vs. offense — who’s winning the AI security race?
Today’s Panel
Jon Swartz, host of Techstrong TV. Mike Vizard, Editor-in-Chief at Techstrong Group. Kate Scarcella, Chief Architect, Cybersecurity & Trusted AI. Sid Nag, VP Analyst at Gartner. Anne Ahola Ward, CEO of CircleClick. Alex Porter, CEO of Lunar Outpost.
Episode Chapters
00:00 — Cold open
01:30 — The OpenAI / Microsoft partnership reset
12:00 — Multi-cloud, IP rights, and the end of Azure exclusivity
20:00 — GitHub Copilot pauses sign-ups: the agentic pricing problem
28:00 — Why $20/month coding assistants don’t survive 2026
35:00 — Project Glasswing: Microsoft embeds Claude Mythos in its SDL
44:00 — The 17-year-old FreeBSD flaw Mythos found alone
52:00 — Defenders vs. attackers: Anthropic’s warning to government
58:00 — Panel predictions and final takes
Read the Full Articles
GitHub Copilot pause announcement: https://thenextweb.com/news/github-copilot-signup-pause-agentic-ai-usage-limits
AI-Powered Defense for an AI-Accelerated Threat Landscape (Microsoft Security): https://www.microsoft.com/en-us/security/blog/2026/04/22/ai-powered-defense-for-an-ai-accelerated-threat-landscape/
Claude Code Security & Project Glasswing (Anthropic): https://www.anthropic.com/news/claude-code-security
CVE-2026-4747 — FreeBSD NFS RCE on NVD: https://nvd.nist.gov/vuln/detail/CVE-2026-4747
Never Miss an Episode
Catch Techstrong Gang live every weekday at 12:00 PM ET, or watch the replay on demand. Subscribe to Techstrong TV on YouTube for daily AI, DevOps, and security analysis. The audio version lands same-day on Apple Podcasts and Spotify.
