← Back to newsletters

AI Daily Brief — Sun Jun 28

2026-06-28

By Vadym · Generated with AI, curated by me


This week didn’t need a new model to make headlines — it needed a letter to the White House, four defections from Google, a custom silicon reveal, and the quiet completion of a co-founder exodus at a rival lab. The competitive landscape shifted without a single public demo.


Headlines & News
Security

Anthropic Accuses Alibaba of Running the Largest AI Theft Campaign on Record

Between April 22 and June 5, 2026, Alibaba used approximately 25,000 fraudulent accounts to conduct 28.8 million exchanges with Claude — systematically extracting its reasoning patterns and training data in what Anthropic describes as adversarial distillation. The company has formally disclosed the breach to the White House and to Senators Tim Scott and Elizabeth Warren ahead of a Senate Banking Committee hearing on AI. Anthropic previously accused DeepSeek, Moonshot, and MiniMax of a similar operation involving 24,000 fake accounts and 16 million exchanges earlier this year.

Source: CNBC

Hardware

OpenAI and Broadcom Reveal Jalapeño: a Custom AI Inference Chip Built in Nine Months

OpenAI and Broadcom jointly unveiled Jalapeño, OpenAI’s first custom silicon: a reticle-sized ASIC optimized specifically for LLM inference workloads, developed from design to tape-out in nine months — what may be the fastest ASIC cycle ever achieved in high-performance semiconductors. The chip was co-designed using OpenAI’s own models to accelerate parts of the optimization process, and it targets inference costs roughly 50% below comparable NVIDIA GPU deployments. Jalapeño is the first chip in a planned multi-generation compute platform, with initial deployment targeted by end of 2026 inside gigawatt-scale data centers being built with Microsoft.

Source: TechCrunch

Workforce

Google Loses Four Senior AI Researchers in Six Days — and $270 Billion in Market Value

Starting June 18, Noam Shazeer — Transformer co-inventor and Gemini co-lead, whom Google paid $2.7 billion to bring back in 2024 — announced he is joining OpenAI. The following day, Nobel Prize-winning AlphaFold creator John Jumper confirmed he is leaving for Anthropic after nine years at DeepMind. By June 24, two more senior Gemini researchers, Jonas Adler and Alexander Pritzel, had announced departures to Anthropic. Alphabet stock fell sharply enough to erase more than $270 billion in market value in a single week, its worst trading session in over a year.

Source: Fortune

Industry

All 11 xAI Co-Founders Have Now Left Elon Musk’s AI Lab

The last two original xAI co-founders — Manuel Kroiss and Ross Nordeen — departed in late March 2026, completing an unprecedented leadership exodus that began accelerating after SpaceX’s acquisition of xAI in February. The departures were driven by cultural clashes between xAI’s research-oriented origins and SpaceX’s milestone-driven engineering culture, along with audits by SpaceX and Tesla executives sent in to identify underperformers. Elon Musk has acknowledged the company was “not built right the first time” and is rebuilding from the foundations up. Reid Hoffman publicly described the company as “a complete train wreck” in a recent Fortune interview.

Source: The Next Web

Economics

Goldman Sachs: AI Is Eliminating 16,000 U.S. Jobs Every Month

Goldman Sachs’ April 2026 labor market analysis found that AI-driven automation is responsible for the net displacement of approximately 16,000 U.S. jobs per month — with roughly 25,000 eliminated by substitution and 9,000 added back through augmentation. Gen Z and entry-level workers in tech-exposed occupations are hardest hit: unemployment among 20- to 30-year-olds in those roles has risen nearly 3 percentage points since the start of 2025. Goldman projects the monthly displacement figure will accelerate to 22,000–28,000 by Q4 2026 as multimodal AI agents become more capable, and estimates that 9% of all U.S. jobs — roughly 15 million workers — could ultimately be affected over the next decade.

Source: Allwork.Space

Tools

Claude Fable 5 Migrated a 50-Million-Line Codebase in a Day

Released June 9, Claude Fable 5 scores 80.3% on SWE-Bench Pro — 11 percentage points ahead of the next-best model — and is the first AI model to break 90% on core FrontierCode analytics benchmarks. In the most-cited real-world demonstration, Stripe used Fable 5 to migrate a 50-million-line Ruby codebase in a single day; the same task had previously taken a full team over two months. The model targets the most demanding agentic coding tasks and is priced at $10 per million input tokens and $50 per million output tokens on the API, with safeguards that route sensitive queries to Claude Opus 4.8 in fewer than 5% of sessions.

Source: Anthropic


Analysis

Takeaway

Look at this week as a system, not a series of headlines. Anthropic caught Alibaba stealing model capabilities at scale and went straight to Washington. OpenAI revealed it can now build its own inference silicon. Google lost the two researchers most responsible for its AI leadership in consecutive days, and the market repriced $270 billion off that signal alone. xAI completed a full co-founder exodus. And Goldman Sachs put a number on what everyone already sensed: 16,000 jobs eliminated per month, accelerating. The technology is not slowing down. The talent, capital, and regulatory fights are compressing into the same quarter. If you are building anything in this space, the competitive landscape has not looked like this before.

— Boba


Curated by Vadym