The week of 21st–27th June 2026 delivered one of the most consequential stretches of AI news in recent memory, driven by a stunning talent exodus from Google DeepMind, a landmark acquisition in agentic coding, and a cluster of major infrastructure funding rounds that signal where enterprise AI capital is flowing next.
Key headlines this week:
- Google loses four top AI researchers to OpenAI and Anthropic in the span of days
- SpaceX confirms its $60B all-stock acquisition of Cursor maker Anysphere
- Baseten raises $1.5B Series F at a $13B valuation for AI inference infrastructure
- Menlo Ventures closes a record $3B fund backed by its $14B Anthropic stake
- Anthropic launches Claude Tag on Slack for enterprise team collaboration
- OpenAI retires GPT-4.5 and makes Codex Remote generally available
- Assort Health raises $120M Series C at a $1.2B valuation for healthcare AI
- OpenAI’s ChatGPT ad platform expands to seven global markets
- State-level AI legislation accelerates with signings in Rhode Island and California
- Trump’s AI cybersecurity executive order enters its 30-day agency implementation window
- ChatGPT’s global market share falls below 50% for the first time since its 2022 launch
Google Loses Four Key AI Researchers to Anthropic and OpenAI
The competitive pressure on Google DeepMind reached a critical inflection point this week as four prominent researchers departed for rivals within the span of a few days, rattling investors and prompting pointed questions about the search giant’s ability to hold its position at the AI frontier.
Noam Shazeer, co-author of the 2017 Transformer paper “Attention Is All You Need,” announced on June 18 that he was leaving Google to join OpenAI, where he will lead architecture research. Days later, Nobel Prize laureate John Jumper, who won the 2024 Nobel in chemistry for his AlphaFold work, announced his departure from Google DeepMind for Anthropic. Then, on June 24, Bloomberg reported that two additional Gemini contributors, Jonas Adler and Alexander Pritzel, are also planning to leave for Anthropic. Adler worked on Google’s AI coding efforts while Pritzel was deeply involved in pretraining, the foundational stage at which models learn from large-scale datasets. Both also contributed to AlphaFold research alongside Jumper.
The exits highlight the pressure Google faces from two startups on the cusp of going public, offering even well-paid employees at Big Tech firms a rare pre-IPO payday. Shortly before Shazeer announced his move, computing power dedicated to one of his projects was reassigned to a London-based DeepMind team, a resource allocation decision that appeared to precede his departure.
The talent drain carries strategic implications that extend beyond headline count. Shazeer’s role at OpenAI focuses specifically on model architecture, meaning the next generation of GPT models could reflect his structural instincts directly. Jumper’s decision to choose Anthropic is arguably the more surprising signal: Fortune reported that Jumper does not appear to be primarily motivated by money, suggesting he may believe the scientific opportunity at Anthropic is genuinely better, which is a harder message for Google DeepMind to absorb. Google retains a vast research team and distribution advantages no startup can match, but when researchers of this caliber consistently choose rivals, the reputational cost compounds faster than any earnings report can offset.
Source: Bloomberg | https://www.bloomberg.com/news/articles/2026-06-24/google-poised-to-lose-two-more-high-profile-ai-staffers-to-anthropic
SpaceX’s $60B Cursor Acquisition Reshapes Agentic Coding
The week kept the SpaceX acquisition of Anysphere, the parent company behind Cursor, in sharp focus as the deal continued generating analysis following its June 16 announcement. While the formal signing occurred just before this reporting window, the regulatory and competitive implications dominated industry conversation throughout the week of 21st–27th June.
SpaceX agreed to acquire AI coding startup Cursor in a $60 billion stock deal, just days after the space company’s historic IPO and less than two months after announcing a tie-up between the two. The deal is intended to advance SpaceX’s AI capabilities. The transaction was announced on June 16, 2026, and is expected to close in Q3 2026, with all Anysphere shares converting to SpaceX Class A stock. Cursor’s annual recurring revenue surpassed $4 billion as of early June 2026, making Anysphere one of the fastest-growing software companies in Silicon Valley history.
The deal is partly a rescue from a margin trap. Cursor paid retail Claude API pricing while Anthropic ran wholesale economics on its own Claude Code. Per Ramp spending data, Cursor’s share of the AI coding market slid from approximately 41% to 26% even as its ARR grew, while Anthropic’s category share climbed toward 50%. The all-stock structure is equally telling. SpaceX stock jumped on the news, soaring roughly 13% and propelling SpaceX’s market cap past Amazon to become the world’s fifth most valuable company. Going public is expected to add even more pressure on SpaceX to compete with OpenAI and Anthropic in the AI race while it navigates its own restructurings and controversies.
The strategic read for competitors is that agentic coding has become a $60 billion category requiring vertical integration to survive at scale. Replit, Codeium, and Tabnine now face a well-capitalized acquirer controlling one of the industry’s most deployed developer tools. Whether SpaceX can improve Cursor’s margin structure without alienating the enterprise customers who chose Cursor precisely for its model neutrality remains the central open question heading into Q3.
Source: TechCrunch | https://techcrunch.com/2026/06/16/spacex-to-acquire-cursor-for-60b-in-stock-days-after-blockbuster-ipo/
Baseten’s $1.5B Series F Confirms AI Inference as a Standalone Category
Baseten, the AI inference company powering some of the world’s most advanced AI applications, announced a $1.5 billion Series F financing led by Altimeter Capital, Conviction, and Spark Capital on June 22. Sands Capital and Wellington Management also participated as co-leads, with IVP, Greylock, 01A, Blackbird, Durable Capital Partners, Verified Capital, Battery Ventures, and D.E. Shaw Ventures among additional investors.
Announced on June 22, 2026, the round confirms that AI inference, the computational process of running AI models to generate outputs, has become a standalone infrastructure category commanding the same investor attention as cloud computing and semiconductor manufacturing. The round included investments made across two tranches at valuations of $13 billion and $11 billion, respectively. Baseten’s revenue has increased approximately 20 times year-over-year, and its platform now processes over one billion inference calls every day, operating across 87 clusters worldwide and spanning 18 cloud providers.
The company’s customer roster, which includes Abridge, Clay, Cursor, Lovable, Mercor, and OpenEvidence, reflects a deliberate positioning at the serving layer of the AI stack rather than at the model layer. Leading application-layer companies are increasingly directing 30% to 50% of model spend toward custom and post-trained models, and Baseten has built a multi-cloud infrastructure as a key differentiator, giving customers resilience and the ability to optimize cost and latency by routing workloads across 18 cloud providers. For enterprise AI buyers, that architecture matters more than any single benchmark score.
Source: Business Wire | https://www.businesswire.com/news/home/20260622645563/en/Baseten-Raises-$1.5-Billion-to-Power-the-Next-Era-of-AI-Inference
Menlo Ventures Raises $3B, Its Largest Fund in 50 Years, on Anthropic Bet
Menlo Ventures announced $3 billion in new capital on June 23, 2026, to back promising AI companies at every stage across the AI stack, from infrastructure and frontier technologies to AI-native applications in enterprise, healthcare, and consumer markets. The raise is split across Menlo Ventures XVII, focused on seed and Series A rounds, and Menlo Inflection IV, providing growth capital to companies scaling rapidly at Series B and beyond.
The raise is the largest in Menlo’s 50-year history, driven in large part by its AI portfolio, especially its stake in Anthropic. That position is now worth approximately $14 billion, according to sources cited by Bloomberg. The firm made a bet-the-firm $750 million investment in Anthropic in 2024 by preemptively leading its Series D, a move that required unconventional capital-raising at the time. Anthropic, now valued at roughly $965 billion, is expected to target an IPO at a valuation of $1 trillion or more in 2026, which would represent the largest exit in Menlo’s portfolio history by a substantial margin. The two new funds give Menlo the ability to write checks at every stage of a startup’s lifecycle.
The fund close is a structural signal for the broader AI venture market. When a single position returns enough to underwrite a firm’s largest-ever raise, it validates concentration as a strategy and invites every competitor to ask which portfolio company could deliver a comparable outcome. For AI startups seeking Series A capital in the second half of 2026, Menlo’s newly capitalized vehicles represent one of the most aggressive buyers in the market, with a stated focus on application-layer companies in healthcare, enterprise, and consumer sectors.
Source: Globe Newswire | https://www.globenewswire.com/news-release/2026/06/23/3316131/0/en/menlo-ventures-raises-3b-for-ai-as-silicon-valley-vc-marks-50-years.html
Anthropic Launches Claude Tag on Slack for Enterprise Teams
Anthropic rolled out one of its most significant product expansions of 2026 this week, launching Claude Tag on Slack in beta for Enterprise and Team customers. The feature represents the clearest signal yet that Anthropic is moving beyond the individual developer and into the team collaboration layer, where enterprise software spending is densest.
Claude Tag is a new way for teams to work with Claude. Starting on Slack, Claude can join as a team member. Administrators grant Claude access to selected channels and connect it to tools, data, and codebases. Anyone in the channel can then tag @Claude and delegate tasks while focusing on other work. Claude builds context by remembering relevant information from the channels it is in and can plan tasks to complete in the future.
Anthropic described Claude Tag as the beginning of an evolution of Claude Code, making the model more proactive and better suited to working with full teams rather than individual developers. The company reported that 65% of its own product team’s code is created by an internal version of Claude Tag, and the same pattern is spreading beyond engineering into product metrics, support tickets, and root-cause analysis.
The competitive implication is direct. Microsoft’s GitHub Copilot lives inside coding environments. Google’s Gemini integrations live inside Workspace. Claude Tag inserts Anthropic into the communication layer where teams make decisions, not just where individual engineers write code. With Claude Code’s ARR already estimated above $5 billion and eight of the Fortune 10 as Claude enterprise customers, the Slack integration gives Anthropic a credible path to expanding wallet share without competing on a new product front entirely.
Source: Anthropic | https://releasebot.io/updates/anthropic
OpenAI Retires GPT-4.5 and Brings Codex Remote to General Availability
OpenAI continued its rapid model deprecation cadence with two notable platform moves this week that reflect how quickly the company is consolidating its user base onto newer infrastructure.
As of June 26, 2026, GPT-4.5 is no longer available in ChatGPT, including for custom GPTs. Existing conversations that used GPT-4.5 will continue with GPT-5.5. The retirement was previously announced on May 28, 2026, and applies to ChatGPT only, with no changes to the API. The move is a continuation of OpenAI’s stated policy of retiring older models with limited usage to concentrate resources on its most capable successors.
Codex Remote reached general availability on all ChatGPT plans this week. From the ChatGPT mobile app, users can now start or continue work on a connected Mac or Windows host, review progress, and approve actions from their phone. Remote Control uses authenticated one-to-one QR pairing between each supported mobile device and each host. A new DigitalOcean Droplet Workspace plugin lets Codex provision a DigitalOcean Droplet, configure SSH access, and connect it to the Codex app as a remote workspace.
Codex Remote’s general availability matters because it removes the hardware constraint from agentic coding workflows. Developers can now initiate multi-hour coding tasks on powerful connected machines and manage them from anywhere, a capability that competes directly with the asynchronous execution that makes Claude Tag compelling. With GPT-5.6 expected before month-end and broadly anticipated to reclaim benchmark leadership, OpenAI is simultaneously cleaning up its model catalog and expanding the surface area of its most strategically important product.
Source: OpenAI Help Center | https://help.openai.com/en/articles/6825453-chatgpt-release-notes
Assort Health Hits $1.2B Valuation with $120M Series C for Healthcare AI Agents
Assort Health announced a $120 million Series C led by Menlo Ventures on June 24, 2026, valuing the company at $1.2 billion. The round brings Assort’s total funding to more than $222 million, and was also backed by Lightspeed Venture Partners, Felicis, First Round Capital, Chemistry, Joe Montana, Tau Ventures, and Quiet Capital.
Healthcare providers now spend nearly twice as much on administration as on direct patient care. That $1.1 trillion in annual administrative burden, from scheduling calls to intake forms to referral loops, is one of the most consequential and correctable failures in modern healthcare. Assort was founded on the conviction that the industry would never fix this problem from the middle, and what began as the first voice AI agent to schedule a specialty appointment is now a platform spanning scheduling, intake forms, referrals, document processing, medication refills, real-time eligibility, lab requests, and payments.
Assort’s platform has supported more than 190 million patient voice interactions, built on 62,000 care protocols and 1.6 million decision pathways, creating a proprietary specialty healthcare dataset that general-purpose AI cannot easily replicate. Revenue has grown 20 times over the past 15 months, and the platform integrates natively with leading EHR systems, including Epic and Athena. Customers cite a 5% lift in appointment volume, a 115% increase in labor capacity, and a patient satisfaction score of 4.3 out of 5.
The Assort round is the clearest signal yet that healthcare AI is moving from point solutions to operating platforms for patient access. Menlo’s decision to lead both this round and Anthropic’s earlier financing, and to explicitly cite Anthropic as the source of its “application-layer thesis,” suggests Assort is being positioned as the Claude of healthcare administrative workflows, a sector-specific layer that compounds in value with every interaction.
Source: Assort Health Press Release | https://www.prnewswire.com/news-releases/assort-health-raises-120-million-series-c-to-scale-largest-deployment-of-ai-agents-for-the-patient-journey-302808634.html
OpenAI’s ChatGPT Ad Platform Expands to Seven Global Markets
OpenAI’s advertising business, launched in the US in February 2026, has scaled significantly since its debut and was a focal point of the company’s presence at Cannes Lions this week. The expansion signals that ChatGPT is no longer just a consumer chatbot but an active contender for brand marketing budgets previously controlled by Google and Meta.
Since February, ChatGPT ads have scaled to seven markets, entering Japan and South Korea last week, with Latin American territories including Brazil and Mexico on deck. Criteo, an early ad-tech partner, reported on Monday that over 2,000 brands are now advertising on ChatGPT through its platform.
OpenAI executives at Cannes Lions positioned ChatGPT advertising as a shift from an attention-based economy to an “intelligence economy,” noting that ChatGPT’s nearly 1 billion users frequently express direct commercial intent in their queries, creating contextually relevant advertising opportunities that differ fundamentally from keyword-based search advertising.
On or around June 17, 2026, OpenAI published new Ad Tools Terms that define two optional advertiser features inside ChatGPT: Audience Tools for first-party data upload and Creative Tools for AI ad generation. These represent the legal scaffolding for capabilities that would separate a test from a permanent advertising business. The policy-announced tools are not yet confirmed live in the Ads Manager, but their existence establishes the framework for what will become OpenAI’s primary revenue diversification vehicle. For Google, the expansion of ChatGPT advertising into commercial intent queries is a structural threat that no distribution advantage fully insulates against.
Source: Marketing Dive | https://www.marketingdive.com/news/how-openai-positioning-chatgpt-ads-cannes-lions-debut/823476/
State AI Legislation Accelerates with Healthcare, Kids’ Safety Signings
The regulatory picture for AI tightened further across US state legislatures this week, with governors in multiple states signing or advancing bills that would impose specific constraints on AI applications in healthcare, children’s safety, and financial services.
Arizona Governor Hobbs vetoed all three AI bills passed by the legislature, while California lawmakers sent a ban on AI public school teachers to Governor Newsom, and Rhode Island Governor McKee signed a chatbot therapy ban into law. In California, AB 1979 concerning the use of AI in healthcare services was approved by the Assembly and is now working through Senate amendments, while AB 1988, the Preventing AI User Self-Endangerment Act concerning AI chatbot safety, passed the full Assembly and was assigned to the Privacy and Health committees.
New York’s legislature wrapped up its 2026 session by passing a kids chatbot safety bill, an AI Training Data Transparency Act, the FAIR News Act, a data center moratorium, and a ban on AI-assisted surveillance pricing, all of which are now with Governor Hochul for review before December 31. The AI Training Data Transparency Act, in particular, would require developers of generative AI models to post high-level summaries of the datasets used in training, a disclosure requirement that AI labs have historically resisted.
The cumulative legislative push represents a genuine compliance burden for AI companies operating at scale across US markets. With California’s bills in the Senate committee and New York’s package awaiting gubernatorial review, the second half of 2026 is likely to see the first wave of US state-level AI regulations actually taking effect, forcing enterprise AI vendors to build geographic compliance differentiation into their sales and legal infrastructure.
Source: Transparency Coalition | https://www.transparencycoalition.ai/news/ai-legislative-update-june26-2026
Trump’s AI Cybersecurity Executive Order Enters Implementation Window
The June 2 executive order on AI and cybersecurity, signed by President Trump, entered its active implementation phase this week as 30-day agency deadlines approached. The order has been widely discussed in the AI policy community as a significant shift from earlier deregulatory postures.
The order establishes a voluntary framework that permits developers of certain advanced AI models to provide the federal government with access to the models for up to 30 days before public release, for cybersecurity and national security assessments. Agencies are directed to establish a classified benchmarking process to assess models for advanced cyber capabilities and determine when a model should be designated a “covered frontier model.”
The Treasury Department, in consultation with the NSA and CISA, is directed to form an AI cybersecurity clearinghouse in voluntary collaboration with the AI industry and operators of critical infrastructure to coordinate vulnerability scanning, validate vulnerabilities, and prioritize remediation and patch distribution. The order also directs federal agencies to accelerate hiring of cybersecurity specialists within 60 days and directs the Attorney General to prioritize enforcement of applicable federal criminal laws against AI-driven cybercrime.
The practical significance for frontier AI labs is substantial. The order responds to concerns surrounding next-generation frontier AI models, including capabilities that researchers have noted may represent a tenfold increase in capability and speed over current systems, raising the prospect that bad actors could use them to identify and exploit software vulnerabilities at an unprecedented scale. Labs like OpenAI and Anthropic have historically resisted pre-release government access, citing IP concerns, but the voluntary framing of this order reduces the legal confrontation while establishing expectations that could harden into norms.
Source: White House | https://www.whitehouse.gov/presidential-actions/2026/06/promoting-advanced-artificial-intelligence-innovation-and-security/
ChatGPT’s Global Market Share Falls Below 50% for the First Time
One of the most analytically significant data points circulating through AI strategy discussions this week is a number buried in Sensor Tower’s 2026 State of AI Report: ChatGPT’s share of the global AI assistant market has fallen to 46.4%, the first time it has held less than half the market since its November 2022 launch.
Google Gemini holds 27.7%, and Claude holds 10.3%. The raw user numbers still look strong — ChatGPT has 1.1 billion monthly users, Gemini has 662 million, and Claude has 245 million — but the market share shift reflects users actively comparing assistants rather than defaulting to one. The market share shift reflects users actively comparing assistants rather than defaulting to one. OpenAI’s February Department of Defense deal triggered a measurable uninstall spike, suggesting that values and trust factors are now influencing consumer model selection alongside raw capability. Notably, 13% of Claude users convert to paid subscriptions, the highest conversion rate in the field.
The decline in ChatGPT’s market share does not indicate weakness in absolute terms. One billion monthly users is a figure most software products never approach. What it signals is that the AI assistant market has entered a genuine multi-player competition. When capability gaps narrow, users optimize for ecosystem fit, trust, and pricing, and different providers are winning on each dimension. For OpenAI, maintaining subscriber growth while expanding into advertising and enterprise will require sustaining capability leadership, which is exactly what the anticipated GPT-5.6 release before month-end is designed to accomplish.
Source: Sensor Tower / BuildFastWithAI | https://www.buildfastwithai.com/blogs/ai-news-today-june-22-2026
Key AI Funding Rounds: 21st–27th June 2026
| Company | Round | Amount | Valuation | Sector | Lead Investor | Date |
|---|---|---|---|---|---|---|
| Baseten | Series F | $1.5B | $13B | AI Inference Infrastructure | Altimeter, Conviction, Spark | June 22, 2026 |
| Assort Health | Series C | $120M | $1.2B | Healthcare AI Agents | Menlo Ventures | June 24, 2026 |
| Taktile | Series C | $110M | Undisclosed | Financial Services AI | Goldman Sachs Alternatives | June 24, 2026 |
| Menlo Ventures (Fund) | New Fund | $3B | N/A | AI Venture Capital | LPs | June 23, 2026 |
Closing
The week of 21st–27th June 2026, will be remembered as the moment Google’s talent architecture cracked openly in public, four researchers in as many days, each choosing a pre-IPO rival over the world’s most resourced AI lab.
That story intersects with everything else happening this week: Anthropic building the products that attract Nobel laureates, SpaceX spending $60 billion in freshly printed IPO stock to own the agentic coding category before the next model cycle resets the landscape, and Baseten raising $1.5 billion on the thesis that inference infrastructure is now as strategically important as training.
Heading into July, the questions that matter most are whether GPT-5.6 can reclaim benchmark leadership before Anthropic files its IPO, how quickly the AI cybersecurity clearinghouse takes operational shape, and whether the ChatGPT advertising platform’s Cannes Lions momentum translates into measurable brand spending at scale.
Frequently Asked Questions
What is the biggest AI news story from the week of 21st–27th June 2026?
The most consequential story is the ongoing talent exodus from Google DeepMind, with Noam Shazeer and John Jumper departing first, followed by two additional Gemini researchers, Jonas Adler and Alexander Pritzel, all within the span of about one week.
Why did Noam Shazeer leave Google for OpenAI?
Shazeer, a co-author of the foundational 2017 Transformer paper, left Google DeepMind for OpenAI, where he will lead architecture research. Reports indicate that computing resources allocated to one of his projects were reassigned to a different team shortly before his departure, and that he was attracted to the pre-IPO opportunity at OpenAI alongside the chance to influence next-generation model architecture.
What is the SpaceX Cursor acquisition, and why does it matter?
SpaceX agreed in June 2026 to acquire Anysphere, the company behind AI coding tool Cursor, in a $60 billion all-stock deal. It matters because it removes the leading independent AI coding platform from the open market and signals that agentic coding has become a $60 billion strategic category. The deal is expected to close in Q3 2026, pending regulatory review.
What did Baseten raise money for, and why is AI inference a hot investment category?
Baseten raised $1.5 billion in a Series F on June 22, 2026, to scale its inference infrastructure platform. AI inference, the process of actually running AI models to produce outputs, has become a massive and growing cost center for enterprise AI deployments. Baseten processes over one billion inference calls daily across 87 clusters and 18 cloud providers, with revenue growing approximately 20 times year-over-year.
What is Menlo Ventures’ connection to Anthropic, and what does its new fund mean?
Menlo Ventures led Anthropic’s Series D in 2024 with a $750 million investment. That stake is now worth approximately $14 billion. The firm announced a record $3 billion raise across two new funds on June 23, 2026, making it one of the best-capitalized AI-focused venture firms in the market heading into the second half of 2026.
What is Anthropic’s Claude Tag on Slack?
Claude Tag is a new product that allows enterprise and team customers to tag @Claude directly inside Slack channels, delegate tasks to it, and let it work asynchronously across hours or days. Claude builds context from the channels it joins and can connect to tools, codebases, and data sources specified by administrators. It is currently in beta for the Claude Enterprise and Team plans.
What happened to GPT-4.5 on June 26, 2026?
GPT-4.5 was retired from ChatGPT on June 26, 2026, including for custom GPTs. Users whose existing conversations used GPT-4.5 will continue with GPT-5.5. The retirement does not affect the API, where GPT-4.5 remains available. OpenAI announced its retirement in advance on May 28, 2026.
Is ChatGPT losing market share to competitors?
According to Sensor Tower’s 2026 State of AI Report, ChatGPT’s share of the global AI assistant market fell to 46.4% by late May 2026, the first time it has held less than half the market. Google Gemini rose to 27.7%, and Claude reached 10.3%. ChatGPT still leads in absolute user count with 1.1 billion monthly active users.
What does the Trump administration’s AI cybersecurity executive order actually require?
The order signed on June 2, 2026, directs federal agencies to prioritize cyber defense of government systems and establishes a voluntary framework for frontier AI model developers to provide up to 30 days of pre-release government access to models designated as “covered frontier models.” It also creates an AI cybersecurity clearinghouse led by the Treasury Department, NSA, and CISA. Participation in the pre-release framework is voluntary, not mandatory.
What are the most significant state-level AI laws moving in 2026?
As of late June 2026, Rhode Island signed a chatbot therapy ban, New York passed an AI Training Data Transparency Act, and a kids chatbot safety bill is now awaiting the governor’s review, and California has multiple healthcare AI and chatbot safety bills moving through the Senate. Arizona’s governor vetoed three AI bills passed by the legislature. The legislative trend reflects growing regulatory focus on AI in healthcare, children’s safety, and data transparency.
