AI News June 28 2026 — It's No Longer OpenAI vs Anthropic: Both Labs Are Under the Same Government Control

1. It's Not OpenAI vs Anthropic Anymore — Both Labs Have the Same Problem

TechCrunch published the clearest editorial verdict of June 2026: the framing of "OpenAI vs Anthropic" has become irrelevant. Both companies are now in exactly the same position with exactly the same problem facing them — and the same disaster waiting if they fail. The US government is set to take an uncomfortable degree of control over which AI models get released, without any formal framework for doing so.

The parallel is now complete. Anthropic's Mythos 5 has been in "preview" for months with no indication it will reach general release any time soon — even after June 27's targeted restoration to critical infrastructure operators. GPT-5.6 Sol/Terra/Luna was restricted to government-approved partners before it ever reached the public, with Altman projecting a "couple of weeks" timeline that rhymes uncomfortably with the language used during the Fable 5 suspension. The Information broke the news that GPT-5.6 would be released only into limited preview, with the government approving the release "customer by customer" until a general release can be approved.

The cost of the current situation is not partisan. It is not something either lab has done to the other. It is a structural problem with applying ad-hoc informal approval processes to products that millions of developers and enterprises depend on. The fix — establishing a formal pre-release review process, as Trump's executive order envisioned but did not implement — is the only path that helps both labs simultaneously. Until that framework exists, every frontier model release from every US lab is subject to the same informal veto. The question for the AI industry in the second half of 2026 is no longer "which lab wins" — it is "who establishes the rules of this new game and how fast."

2. Anthropic Science Event June 30 — What to Expect Monday

"The Briefing: AI for Science" streams live Monday June 30 at 10am PST. The event brings together Anthropic leadership, pharma executives, biotech founders, and leading research institutions. It is the most substantive AI-for-science event Anthropic has ever hosted — and it arrives at a moment when the company has spent six months building the infrastructure to justify the positioning.

The research context matters. In June 2026, Anthropic published "Paving the Way for Agents in Biology," which introduced VirBench — 120 viral sequence retrieval queries across 40 pathogens. The findings were stark: Claude Sonnet 4 achieved just 16.9% accuracy on identical viral sequence queries across repeated runs using standard retrieval. After Anthropic collaborated with NCBI to build a deterministic retrieval tool (gget virus), accuracy jumped to 92.8%. GPT-5.5 went from 91.3% to 99.7% on the same benchmark. The lesson: reliable AI agents in science require deterministic data infrastructure, not just better models. A cheaper model with the right tool beat expensive models without one.

What to watch on Monday: product demonstrations from Anthropic leaders, customer showcases from pharma and biotech (Bristol Myers Squibb is already deploying Claude across R&D and manufacturing), and — strongly expected — John Jumper's first public appearance at Anthropic. The Nobel laureate who co-created AlphaFold joined Anthropic weeks ago. June 30 is the natural moment for his first public role statement. Anthropic has built biology benchmarks, wet lab infrastructure, and partnerships with the Allen Institute and HHMI. Jumper is the person who proved AI can do science at Nobel scale. The event should reveal what that combination produces.

Related: John Jumper joins Anthropic — full analysis · Google loses 4 AI researchers to Anthropic in one week

3. Claude Opus 4.7 Fast Mode Deprecated July 24 — Migrate Now

Anthropic deprecated fast mode for Claude Opus 4.7, with removal on July 24, 2026. After removal, requests to claude-opus-4-7 with speed: "fast" will return an error. The migration path is straightforward: switch to fast mode for Claude Opus 4.8, which carries the same fast mode capability with improved performance across coding, agentic tasks, and professional work. Any production system still using Opus 4.7 fast mode has 26 days to migrate before requests start failing.

The Opus 4.7 deprecation continues the pattern of Anthropic consolidating its model lineup following the May 28 Opus 4.8 launch. Fast mode for Opus 4.7 has been available since 2025 — its removal is a routine deprecation, not a capability loss, since Opus 4.8 supersedes it on every dimension. Developers on Claude Code and direct API should audit their model strings this weekend.

4. Reid Hoffman: xAI Is a "Complete Train Wreck" — Cursor May Have Already Peaked

LinkedIn co-founder and Anthropic investor Reid Hoffman made some of his most pointed public comments about the AI competitive landscape in a Fortune interview. On xAI: all 11 of xAI's original co-founders had departed the company by May 2026 — a cascade that began in earnest in February when Tony Wu, described as one of the most operationally central co-founders, announced his resignation. Hoffman's read on SpaceX acquiring Cursor for $60 billion: "You could almost think of it as the IAC of AI" — a serial acquisitions roll-up strategy built on aggregation rather than organic capability. He called SpaceX "not an AI company" and xAI "a complete train wreck."

On Cursor specifically: "Cursor seems to have had its bright star some number of months ago and seems to be fading over the horizon." Cursor has faced mounting pressure since early 2026 as Claude Code and Codex gained ground. SpaceX acquired Anysphere (Cursor's parent) at $60 billion in all-stock on June 16 — before the SpaceX stock dropped 16% in one session. The Cursor acquisition at peak SPCX valuation, followed by the stock collapse, has raised questions about whether the deal's effective value has already eroded significantly for Anysphere shareholders. See the full SPCX analysis: SPCX -16% — what happened →

Hoffman's competitive framing for the field: Anthropic strong in code and expanding into design and legal. ChatGPT functioning as a consumer search front end, with Codex "insufficiently talked about given its strength." His framing of xAI is pointed but he is not a neutral observer — Anthropic is a portfolio company and he has financial interest in its competitive positioning.

5. Goldman Sachs: AI Erasing 11,000 US Jobs Per Month

Goldman Sachs's semi-regular AI jobs tracker found in June 2026 that AI was erasing approximately 11,000 net US jobs per month — down slightly from 16,000 per month in April 2026 — with Gen Z bearing a disproportionate share of the impact as entry-level knowledge roles face the highest displacement risk. Separate research found graduate unemployment had risen from 3.6% in 2019 to 5.6% in 2026. By mid-2026, 35% of entry-level job postings required at least three years of experience, and 45% of companies were using automated rejection systems at early hiring stages.

Hoffman's counter-argument on the jobs data: most of the entry-level slump is being misattributed to AI when it has more to do with "global turbulence and businesses being unable to figure out how to invest and plan." He characterized the AI-washing of layoff decisions as companies using AI as a convenient narrative for decisions driven by other factors. The debate will not be resolved by June 2026 data alone — but the Goldman Sachs tracker's consistent 10,000-16,000 monthly net job loss figure is the hardest quantitative evidence available.

What to Watch This Week