AI News Roundup — June 13, 2026
The White House's decision to force Anthropic's Fable 5 and Mythos 5 offline was driven by intelligence that a China-linked group may have accessed Mythos — and Amazon's Andy Jassy reportedly tipped off the administration about a jailbreak that Dario Amodei refused to patch. Meanwhile, SpaceX closed its first full trading day up 25%, MSCI forced buying began Saturday morning, and Anthropic quietly overtook OpenAI in US business adoption for the first time.
White House Export Controls on Anthropic Linked to Chinese Access to Mythos — Amazon Tipped Off Government
New reporting from Semafor reveals the real reason behind Friday's forced shutdown of Claude Fable 5 and Mythos 5: the White House received intelligence that a China-linked group may have accessed Mythos, raising fears about reverse-engineering and "distillation" of the model's capabilities. The Trump administration was also warned by Amazon — whose CEO Andy Jassy personally contacted administration officials — that Fable 5 contained a jailbreak capable of bypassing its cybersecurity and biology safeguards. Trump AI adviser David Sacks said that when the administration notified Anthropic about the jailbreak, CEO Dario Amodei "said the jailbreak was not a serious risk and refused to fix it," prompting the export control directive. Anthropic pushes back: the company said the White House never raised Chinese access concerns during their conversations, and that it "prohibits access to its products from within China." It maintains the jailbreak was "narrow" and that evidence was only provided verbally. Sacks' response: "In the past, Anthropic has always said safety must be top priority. In this case, Anthropic prioritized the continued offering of the consumer model over safety." He added: "The ball is in Anthropic's court."
SpaceX Closes Up 25% on First Full Trading Day — MSCI Forced Buying Begins Saturday
SpaceX (SPCX) closed its first day of trading at $168.70, a 25% gain over the $135 IPO price, with an intraday range of $150.00–$176.52. The opening price of $150 — exactly at the analyst threshold for "orderly debut" — signaled market endorsement of 90x EBITDA multiples for AI infrastructure companies, clearing the way for Anthropic and OpenAI to proceed with their own listing timelines. MSCI began adding SPCX to its large-cap index products on Saturday June 13 — the first trading day after the IPO — creating structural forced buying from passive index funds proportional to their benchmark weightings. A second wave of mandatory buying is expected around early July when SPCX becomes eligible for Nasdaq-100 inclusion after 15 trading days. Wall Street registered at least 25 SPCX-linked ETFs before a single share traded, more than half of them leveraged products — an unusual number that signals issuers expect significant short-term volatility.
CNBC · Build Fast With AI · Kiplinger
Anthropic Overtakes OpenAI in US Business Adoption for the First Time
The May 2026 Ramp AI Index — tracking spending patterns across more than 50,000 US businesses — shows Anthropic's business adoption reached 34.4% of companies in April 2026, surpassing OpenAI's 34.4% for the first time in the AI race. Anthropic rose 3.8 percentage points while OpenAI fell 2.9 points to 32.3%. Overall AI adoption among tracked businesses hit 50.6% — a majority of US companies now paying for AI tools. The engine behind Anthropic's growth: Claude Code, now the fastest-growing product in the company's history. One analysis estimates 4% of all public GitHub commits worldwide were authored by Claude Code as of the most recent measurement — double the share from one month prior. However, a separate IDC survey complicates the picture: only 19% of 1,000+ enterprise organizations report "extensive" use of Claude, versus higher rates for OpenAI and Google. The Ramp data measures adoption breadth; the IDC data measures depth. Anthropic leads on the former, trails on the latter.
Build Fast With AI · VentureBeat
China Humanoid Robot IPO Wave Accelerates — EngineAI Files Confidentially for Hong Kong Listing
Chinese humanoid robot maker EngineAI filed confidentially for a Hong Kong IPO, joining a rapidly expanding wave of Chinese robotics companies racing to public markets. Sector leader Unitree has filed for a $7 billion IPO. BYD-backed hand-maker PaXini Tech is also weighing a listing. Robot vacuum giant Dreame is eyeing Hong Kong, and robot-hand unicorn Linkerbot is pursuing a $6 billion valuation. Approximately $22.6 billion has already been raised in Hong Kong IPOs this year, much of it in AI and robotics. The wave reflects both the maturation of China's robotics manufacturing base — which already accounts for ~85% of global humanoid shipments — and investor appetite to participate in a sector that Chinese policymakers have designated a national strategic priority alongside AI.
The Next Web · Build Fast With AI
GPT-5.6 Checkpoint Leaks in Developer Channels — Most Compressed Frontier Release Window Ever
A leaked GPT-5.6 checkpoint is circulating in developer channels, adding to a model release cadence in June 2026 that has no modern precedent. Within a single month, the industry has seen Claude Fable 5 and Mythos 5 (launched and then pulled), Microsoft MAI Thinking One (unveiled at Build), Alibaba Qwen 3.7 Plus, and a still-pending Gemini 3.5 Pro (announced at Google I/O with a "give us until next month" commitment from Sundar Pichai). Polymarket pricing had GPT-5.6 at 80–89% probability of release. The compression of frontier model releases into simultaneous windows is forcing developers and enterprises to make infrastructure and vendor decisions under conditions of extreme uncertainty — the practical guidance from analysts is unchanged: make roadmap decisions based on shipped models, not leaks.
Build Fast With AI · Geeky Gadgets
Anthropic's Troubled Partner Relationships — Surprise Pricing Changes and Competitive Blindsides
The Information published a detailed account of how Anthropic has strained relationships with some of its business partners through sudden competitive moves and pricing changes. Key pattern: Anthropic has launched products that directly compete with partner business lines without advance notice — a practice that has caused friction with companies that had built offerings on top of Claude. Separately, Anthropic switched Claude Enterprise customers from flat per-seat fees to usage-based billing with little warning, a change one analyst estimated could double or triple costs for heavy users. Combined with the abrupt Fable 5 shutdown — which cut off paying customers mid-session — a picture is emerging of a company that has grown faster than its enterprise relationship management practices. The timing is problematic: partner trust is a key enterprise IPO asset, and these dynamics will be scrutinized by public market investors evaluating Anthropic's long-term revenue durability.
The Information · Build Fast With AI
SpaceX's $41.3 Billion Accumulated Loss — The History Behind the $1.77 Trillion Valuation
With SPCX now a public company, the S-1's disclosure of a $41.3 billion cumulative loss since SpaceX's founding in 2002 is drawing attention. The figure is not unusual for a capital-intensive aerospace company in early rocket development stages, and Starlink's emergence as a genuinely profitable segment — $11.4 billion in 2025 revenue, 63% EBITDA margin — represents the turnaround that makes the IPO timing credible. But it is a useful reminder that SPCX at $1.77 trillion reflects a bet on future Starlink and xAI economics, not historical profitability. The entire AI bull case — the $2.17 billion per month in compute contracts from Anthropic and Google — is built on contracts terminable with 90 days' notice after December 31, 2026. Investors buying in the open market are underwriting a forward thesis with significant execution risk on multiple fronts.