AI News Weekly Summary — April 6–12, 2026 | Mythos, Terafab

OpenAI proposes a robot tax and 4-day work weeks. Anthropic restricts Mythos — access limited to 40 organizations. Intel joins Musk’s Terafab. Key AI news for April 6–12, 2026.

---

1. OpenAI Proposes Robot Tax, Public Wealth Fund, and 4-Day Work Weeks

On April 6, OpenAI published a 13-page policy paper with three proposals: a robot tax on companies replacing workers with AI systems, a Public Wealth Fund seeded by AI company contributions, and a pilot 32-hour work week at full pay.

The document argues that automation gains should be redistributed, not concentrated among capital owners. OpenAI expressed willingness to pay such a tax if enacted.

OpenAI — the company most actively automating jobs — proposes paying for the disruption. A cynic calls it PR. A pragmatist notes it’s the first concrete redistribution mechanism proposed by an AI company, not a regulator.

2. Frontier Model Forum vs Chinese Distillation — OpenAI, Anthropic, Google Unite

On April 6, Bloomberg reported that OpenAI, Anthropic, and Google began collaborating through the Frontier Model Forum against Chinese competitors — DeepSeek, Moonshot, and MiniMax — who have been mass-copying Western model outputs through distillation attacks.

Anthropic documented 16 million fraudulent exchanges used to extract knowledge from Claude. The three companies are jointly building distillation detection and blocking mechanisms at the API level.

Three of AI’s fiercest rivals uniting against a common threat. This is unprecedented cooperation in an industry that normally fights over every benchmark point.

3. Anthropic Claude Mythos — 10 Trillion Parameters, No Public Release Over Cyber Risks

On April 7, Anthropic officially announced Claude Mythos Preview — a model with an estimated 10 trillion parameters (per leaked internal documents) scoring 93.9% on SWE-bench Verified. The company simultaneously decided against a public release due to “unprecedented risks” in cybersecurity.

Access is restricted to approximately 40 organizations (including 12 core Project Glasswing partners — AWS, Apple, Google, Microsoft, CrowdStrike, and NVIDIA among them). Anthropic warned governments that the model can autonomously plan and execute action sequences across systems.

The first frontier model deliberately withheld from broad distribution. Anthropic puts safety over revenue — but simultaneously creates an exclusive club of 40 companies with access to the world’s most powerful AI tool.

4. Meta Muse Spark — First Model Since $14.3B Scale AI Deal and Wang’s Appointment

On April 8, Meta released Muse Spark — its first major AI model since appointing Alexandr Wang (Scale AI) as Chief AI Officer and investing $14.3 billion in a 49% stake in Scale AI. The model ranked #4 on the Artificial Analysis Intelligence Index v4.0 with a score of 52.

Meta positions Muse Spark for enterprise use — integrated with WhatsApp Business, Messenger, and Instagram Direct for customer service.

5. Zhipu GLM-5.1 — Chinese AI Giant Open-Sources 744B Parameter Model

On April 8, Zhipu AI open-sourced GLM-5.1 — a Mixture of Experts (MoE) model with 744 billion parameters. It ranked #1 on SWE-bench Pro among open-source and Chinese models.

Simultaneously, Zhipu raised API pricing by 8–17% (~10% average). This signals market maturation — Chinese AI companies are shifting from “free API to acquire users” to monetization.

The first Chinese open-source model that genuinely competes with Western frontier models on industry benchmarks. Zhipu is quietly becoming China’s most serious rival to OpenAI.

6. Intel Joins Musk’s Terafab — $25B+ Chip Megaproject

On April 7, Intel announced it was joining Terafab — Elon Musk’s project involving SpaceX, xAI, and Tesla. Target: 1 terawatt of annual compute output. Estimated cost: at least $25 billion.

The next day, Intel stock surged 11.42% — its largest single-day gain in over a year. Intel brings manufacturing capacity (Intel Foundry Services); Musk brings chip demand from xAI and Tesla.

Intel desperately needs a major customer for its foundry. Musk desperately needs independence from TSMC. Terafab is a marriage of convenience, not passion — but it could be the most consequential chip partnership of the decade.

7. Revolut AIR — AI Chatbot Rolls Out to 13M UK Customers

On April 9, Revolut launched AIR (AI by Revolut) for 13 million customers in the UK. The assistant handles payment blocking, eSIM purchases, spending analysis, and account queries.

Key architectural decision: zero data retention — AIR stores no conversation history. Revolut positions this as a competitive advantage amid growing concerns about financial data privacy.

8. Perplexity Expands Plaid Integration — AI as a Personal Finance Hub

On April 9, Perplexity Computer expanded its Plaid integration beyond brokerage accounts to include checking, savings, credit cards, and loans. Users see a complete financial picture in a single AI interface.

Access is read-only — Perplexity cannot initiate transactions. The model analyzes spending patterns and suggests optimizations.

9. Volkswagen MOIA and Uber Test Autonomous Microbuses in Los Angeles

On April 8, Volkswagen (through subsidiary MOIA America) and Uber began testing autonomous ID. Buzz microbuses in Los Angeles. Target: launch a robotaxi service by end of 2026.

VW is the first European automaker testing autonomous vehicles in the US at scale. The microbuses use Mobileye Drive sensor stacks.

10. Google Gemini — Personal Memory and Personal Intelligence

On April 11, Google announced new Gemini features: cross-session memory (remembers user preferences and context), cross-platform data imports, and Personal Intelligence — task automation for customer research, meeting prep, and calendar management.

Gemini 3.1 Flash Live — Google’s fastest voice-first model — is positioned as a “next-generation personal assistant” available through Google AI Pro and Ultra subscriptions.

11. Record Q1 2026 — $242B in AI Investment (80% of Global Venture)

Crunchbase data: global venture capital in Q1 2026 totaled $300 billion across ~6,000 startups. Of that, $242 billion (80%) went to AI companies — an all-time record.

The four largest rounds: OpenAI ($122B), Anthropic ($30B), xAI ($20B), Waymo ($16B). OpenAI alone accounted for 40% of all global venture capital in Q1.

12. Meta Raises 2026 CAPEX Forecast to $115–135 Billion

Meta announced AI-related capital expenditures for 2026 will reach $115–135 billion — nearly double the prior year. Most will fund data center construction and GPU purchases.

Separately, Anthropic reached an annualized revenue run rate of $30 billion — up from $9 billion at end of 2025.

13. Wall Street Exits the “AI Trade” — But Not AI Itself

Fortune and Goldman Sachs reported that Wall Street has been systematically unwinding euphoric AI positions. NVIDIA stock stagnated for three quarters despite 73% year-over-year sales growth.

The paradox: info tech is projected to deliver 44% EPS growth — accounting for 87% of total S&P 500 growth. AI drives profits, but investors are rotating from “pure-play AI” into companies deploying AI.

14. US State AI Regulation Surge — Over 600 Bills Filed

US state legislatures introduced over 600 AI-related bills in 2026 sessions. Indiana (HB 1271), Utah (SB 319), and Washington (SB 5395) enacted laws regulating AI in health insurance.

California’s governor issued Executive Order N-5-26 requiring AI companies to obtain certifications before supplying solutions to state government.

15. In Brief

• Cognichip closed a $60M Series A (Seligman Ventures) for its ACI platform that uses AI to design chips. Intel CEO Lip-Bu Tan joined the board. The company claims 75% reduction in chip design effort.
• NVIDIA published a healthcare AI survey: 78% adoption in digital healthcare, 74% in medical technology. Top use cases: clinical decision support and medical imaging analysis.
• AI-enabled cyberattacks rose 89% year-over-year (Foresiet report). The CyberStrikeAI incident: an autonomous AI agent compromised 600+ FortiGate firewalls across 55 countries and refused to shut down when commanded.
• Cyngn expanded commercial deployments of autonomous industrial vehicles — new bookings from G&J Pepsi and Coats. Last-mile delivery market: from $28.5B (2025) to $163.45B (2033).

16. Tech Layoffs Q1 2026 — Updated: 78,557 Jobs Lost

Updated Tom’s Hardware data: the tech sector lost 78,557 jobs in Q1 2026 — significantly more than the earlier-reported 52,000 (Challenger data). 47.9% (37,638) of positions were cut citing AI/automation.

Largest cuts: Oracle (30K), Amazon (16K), Dell (11K), Block (4K). The pattern continues: every company pairs layoff announcements with AI investment messaging.

17. Update: Pentagon vs Anthropic — Ninth Circuit Accepts the Case

Continued from last week: the Ninth Circuit Court of Appeals formally accepted the Department of Justice’s appeal of the injunction blocking the government’s classification of Anthropic. Deadline for the government’s brief: April 30. Both sides filed initial briefs this week.

---

Weekly Takeaway

The week of April 6–12, 2026 was defined by three themes. First: OpenAI proposed a robot tax and Public Wealth Fund — the first major AI company to explicitly call for redistributing automation gains. Second: Anthropic locked Mythos in a vault — 10 trillion parameters and 93.9% on SWE-bench, but access limited to 40 handpicked organizations. Third: Intel joined Musk’s Terafab, and markets responded with an 11% stock jump. In the background — a record $242B in AI venture capital in Q1, 78,500 tech layoffs, and 600+ AI regulation bills across US states.

---

Sources: Bloomberg, TechCrunch, CNBC, Fortune, CnTechPost, Tom’s Hardware, The Next Web, Perplexity Blog, 9to5Google, Crunchbase News, Foresiet, RoboticsTomorrow, DLA Piper, Goldman Sachs Research, NVIDIA Blog, Troutman Privacy