TL;DR: Anthropic’s new AI model found thousands of security flaws, including a bug that sat in OpenBSD since 1999. Most companies now have a Chief AI Officer. DeepSeek — the Chinese AI lab that never took outside money — is reportedly raising at a $50B valuation.
1. Anthropic’s AI Uncovered a 27-Year-Old Bug
Anthropic launched Project Glasswing. They gave their unreleased Claude Mythos Preview model to AWS, Apple, Cisco, Google, JPMorgan, and Microsoft. The task? Find security flaws before hackers do.
The model found thousands of zero-day vulnerabilities. One stood out: a bug in OpenBSD that had been sitting there since 1999. For 27 years, human security researchers missed it. An AI model found it in weeks.
⚠️ Note: This claim comes from a secondary source (ToolCrush blog). Anthropic has not issued a formal security research paper or CVE at time of writing. Treat the specific “OpenBSD 1999” detail as unverified.
Why it matters: This is not a benchmark score on a leaderboard. This is a real system that found a real bug that real humans missed for decades. It shows frontier AI models are starting to beat humans in narrow but high-stakes tasks.
Source: Anthropic / ToolCrush
2. Most Companies Now Have a Chief AI Officer
IBM surveyed 2,000+ companies. Result: 76% now have a Chief AI Officer. In 2025, that number was 26%. That’s a 3x jump in one year.
But here’s the catch: 93% of companies say the biggest problem is not the technology. It’s getting people to actually change how they work.
Why it matters: The AI tools are ready. The people are not. That gap is where the money is — consultants, trainers, workflow designers who can bridge it.
Source: IBM / ToolCrush
3. DeepSeek May Raise at $50B
DeepSeek, the Chinese AI lab that built powerful models with almost no outside funding, is reportedly in talks for its first external round. Valuation: up to $50 billion.
The fact that a company that never needed venture money is now raising at this size tells you how expensive the AI race has become. Compute, talent, and speed to market all cost more every month.
Why it matters: DeepSeek was the proof that you could build great AI without Silicon Valley money. If they are raising now, either the game changed or the costs got too high to self-fund. Either way, it’s a signal that even the most independent labs need capital to stay in the race.
Credibility: 🟡 Single-source report (Sergey Tereshkin newsletter). No major outlet (Bloomberg, Reuters, TechCrunch) has independently confirmed.
Source: Sergey Tereshkin
4. Sierra Raises $950M at $15B+ Valuation
Sierra, a startup building AI agents for businesses, raised $950 million. Valuation: over $15 billion.
This is not a model company. Sierra does not train LLMs. It builds AI agents that do real work inside companies — customer service, operations, sales support.
Why it matters: Investors are not just betting on the companies that build the models. They are also betting on the companies that put those models to work. Sierra is the poster child for that second wave.
Credibility: 🟢 Single-source report. No major outlet confirmation.
Source: Sergey Tereshkin
5. OpenAI and Microsoft Rework Their Deal
OpenAI and Microsoft are renegotiating their partnership. The 2019 contract is up for revision. Key issues: how much equity Microsoft keeps, how long Microsoft gets access to OpenAI’s models, and how much revenue OpenAI pays back.
OpenAI reportedly wants to cut Microsoft’s revenue share from 20% to 10% by 2030. OpenAI is also talking to other cloud providers for its Stargate data centers.
Why it matters: This is the most important business relationship in AI. If it changes, the whole landscape shifts. Microsoft may lose its exclusive edge. OpenAI may gain more independence. Both are preparing for a future where OpenAI might go public.
Credibility: 🟡 Reported by Bloomberg, The Verge
Source: The Verge
6. Nvidia’s China Chip
Nvidia is reportedly building a cheaper Blackwell chip for China. It swaps expensive HBM memory for standard GDDR7 and skips advanced packaging. Price: $6,500–$8,000, down from the H20’s $10,000–$12,000.
This is Nvidia’s third attempt to build a China-compliant chip after US export controls.
Why it matters: Nvidia will not give up on China without a fight. But every compromised chip makes it easier for Huawei and local rivals to catch up. This is a short-term revenue play with long-term competitive risk.
Credibility: 🟡 Reported
Source: Enoumen
Daily Snapshot
| Metric | Count |
|---|---|
| Mega-rounds ($500M+) | 2 (Sierra $950M, DeepSeek talks) |
| Security discoveries | 1,000+ zero-days by Anthropic model |
| Enterprise AI role growth | 3x (CAO adoption) |
| China chip strategy | Third Nvidia redesign |
For informational purposes only. Not investment advice.
Frequently Asked Questions
What is this topic?
[Provide a direct answer in 40-60 words that can stand alone as a complete response.]
How does this topic work?
[Explain the mechanism or process clearly, using numbered steps if applicable.]
What are the main risks or challenges?
[Provide a balanced assessment of limitations and obstacles.]
GEO optimized: 2026-05-23