May 29 Roundup: Anthropic Nears $1 Trillion, Claude Opus 4.8 Ships, Illinois Leads on AI Safety, OpenAI Clears the Model Deck, and Google's AI Search Backlash
Thursday delivered a cascade of industry-defining news: Anthropic vaulted past OpenAI to become Silicon Valley's most valuable AI startup at a staggering $965 billion valuation — and simultaneously released its sharpest model yet. Meanwhile Illinois drew a new line on AI safety that Washington refuses to draw, OpenAI quietly announced sunset dates for two of its most-used models, Europe extended its compliance clock by 16 months, and Google's sweeping AI Search overhaul is driving a measurable migration to privacy-first alternatives. Here's everything you need to understand about where AI stands this morning.
1. Anthropic Closes $65B Series H at a $965 Billion Valuation — and Ships Opus 4.8 the Same Day
In what may be the defining capital event of the AI era so far, Anthropic announced Thursday that it has closed a $65 billion Series H funding round at a post-money valuation of $965 billion — eclipsing rival OpenAI (last valued at $730 billion) to become Silicon Valley's most valuable private AI company. The round was co-led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, with institutional investors including Baillie Gifford, Blackstone, Brookfield, D.E. Shaw, DST Global, and Fidelity joining in. Strategic infrastructure partners Samsung, SK Hynix, and Micron also participated. A portion of the round — $15 billion — is made up of previously committed investments from hyperscalers, including a $5 billion Amazon contribution announced in April.
"Anthropic has closed a $65 billion Series H round at a $965 billion post-money valuation, marking what could be the AI startup's last private fundraising before debuting on the public markets."
The company says it will use the fresh capital to "advance our safety and interpretability research, expand compute to meet growing demand for Claude, and scale the products and partnerships our customers rely on." Critically, the round came on the same day Anthropic released Claude Opus 4.8, its new flagship model touting significantly better capabilities in agentic tasks, advanced coding, and a renewed focus on honesty and self-correction. According to reporting from Bloomberg, Anthropic is also reportedly planning to more broadly release models on par with its powerful Mythos cybersecurity model — which has, until now, been distributed only in limited fashion due to potential safety concerns.
The optics are deliberate: Anthropic paired its biggest capital raise ever with a new flagship model release on the same news cycle. The message is unambiguous — Anthropic isn't just chasing OpenAI's valuation, it's trying to lap it on both the safety narrative and the product roadmap simultaneously.
A $965 billion valuation for a five-year-old AI company is a number that demands context. Anthropic's revenue trajectory, particularly the reported $10.9 billion Q2 run rate, gives the number some grounding — but the real story is strategic positioning ahead of an IPO that looks increasingly inevitable. For enterprise buyers, this raise changes the calculus: Anthropic is no longer the "safer but smaller" alternative to OpenAI. It is now the better-capitalized one. Expect enterprise sales cycles to shorten significantly as procurement teams process the signal that Anthropic has the financial runway to compete indefinitely. The simultaneous Opus 4.8 launch is also notable: Anthropic has figured out how to time its capital and product news for maximum narrative impact.
2. Illinois Signs Landmark AI Safety Law as Washington Steps Back
In a sharp pivot from federal inaction, Illinois Governor J.B. Pritzker signed SB 315 into law this week, positioning the state at the forefront of AI safety regulation nationwide. The legislation came just days after President Donald Trump unexpectedly shelved a federal plan aimed at regulating frontier AI models, citing concerns over stifling innovation — a move that opened the door for state-level action.
SB 315 imposes substantive requirements on major AI companies operating in Illinois: firms must submit detailed public safety plans and annual reports summarizing third-party safety evaluation outcomes for their most advanced models. Additionally, companies are required to report any significant safety incidents to the state within 72 hours — or within 24 hours if an incident poses an imminent risk of death or serious bodily harm. The law also creates protected whistleblower channels for AI company employees to surface safety concerns.
"A few days after President Donald Trump unexpectedly scrapped a federal plan aimed at regulating frontier AI models, Illinois passed a law that could set a new precedent in AI safety regulation nationwide."
— Legal News Feed, May 28, 2026
Governor Pritzker emphasized Illinois's leadership role in holding large technology firms accountable, asserting that the state would not shy away from pioneering robust safety measures even in the absence of federal guidance. The law's passage raises serious questions about the emerging patchwork of state-by-state AI regulations that tech firms will need to navigate. Illinois follows California, which has already introduced stringent AI procurement guardrails, and Spain, which this month approved a draft Organic Law on AI governance aligned with the EU AI Act framework.
Illinois SB 315 isn't just a local compliance story — it's a signal that state-level AI regulation is filling the vacuum left by federal inaction. The 72-hour incident reporting requirement alone is operationally significant: it means AI companies need internal incident classification and escalation workflows that rival those in financial services and healthcare. For enterprises deploying AI in Illinois, this raises due diligence requirements for your AI vendors: can they demonstrate their safety plans meet SB 315's public disclosure standard? If your AI vendor is a major frontier lab, ask them directly. If your vendor is a smaller integrator, they're likely relying on upstream model providers — and you need to understand where responsibility sits in that chain.
3. OpenAI Announces Retirement Dates for o3 and GPT-4.5 as Model Lifecycle Discipline Arrives
OpenAI this week quietly updated its Model Release Notes to announce the retirement of two of its most prominent recent models: GPT-4.5 will be sunset from ChatGPT on June 27, 2026 (following a 30-day sunset period), and o3 will be retired on August 26, 2026 (following a 90-day sunset period). The announcements reflect a broader strategic shift at OpenAI — the company is actively consolidating its model portfolio around its newer GPT-5.5 family and clearing the slate of transitional models that served as stepping stones.
The move comes in the context of OpenAI's increasingly crowded model catalog. The company has been shipping models at a high cadence throughout 2026 — from GPT-5.5 Instant to GPT-5.5 Pro to ChatGPT Images 2.0 — and the accumulation of overlapping offerings has created confusion for enterprise buyers trying to standardize their integrations. The retirements signal that OpenAI is now serious about model lifecycle hygiene, treating its model roster more like a SaaS product roadmap and less like an experimental research catalog.
"OpenAI o3 will be retired from ChatGPT on August 26, 2026 following a 90-day sunset period, and GPT-4.5 will be retired from ChatGPT on June 27, 2026 following a 30-day sunset period."
— OpenAI Help Center, May 28, 2026
For enterprise API users, the timelines are tight — particularly the June 27 date for GPT-4.5. Any production systems calling gpt-4.5 directly will need migration plans in place within days, not weeks. OpenAI has historically provided migration paths to successor models, and it is expected that GPT-5.5 will serve as the functional replacement for both retired models in most use cases.
If your engineering team has any production workloads on GPT-4.5, this is a five-alarm fire on your sprint board. June 27 is weeks away. Model migrations sound routine, but they frequently surface subtle prompt sensitivity differences — responses that were calibrated for GPT-4.5's specific output style may behave differently on GPT-5.5. Build in buffer time for evaluation and regression testing. More broadly, the era of "pick a model and forget it" is definitively over. Every enterprise AI deployment needs a model lifecycle management strategy. OpenAI is now forcing that conversation with hard deadlines — and it won't be the last time.
4. Google's AI Search Overhaul Triggers a DuckDuckGo Surge — and Questions About Search's Future
The ripple effects of Google's sweeping AI Search announcements at I/O 2026 are now measurable and significant. Following Google's May 19 rollout of deeper AI features into the core search experience, DuckDuckGo reported that US installs rose an average of 20.8% week-over-week, peaking at 37.6% growth on May 26. On iOS in the US, installs climbed an average of 33%, reaching nearly 70% growth on May 25. Visits to DuckDuckGo's noai.duckduckgo.com page — where AI features are disabled by default — rose 22.7% week-over-week.
"Google is force-feeding AI with no way to opt out. Some users are pushing back on that approach."
— Business Insider / Alistair Barr, May 28, 2026
The context: at I/O, Google announced a redesign of Search that integrates AI Mode capabilities — including conversational long-form queries, multimodal inputs (images, videos, files, browser tabs), and AI-generated answers — directly into the main search box. The changes also introduced Gemini 3.5 Flash as the default model for the Gemini app and AI Mode in Search, and launched Gemini Omni, a new multimodal model family capable of generating video clips from text, photo, video, and audio prompts. The additional AI agent Gemini Spark runs 24/7 in the background, connected to Google Workspace.
Google's I/O announcements were also notable for Project Aura smart glasses updates — a clear shot across Meta's Ray-Ban smart glasses bow — and an expanded role for Gemini Omni across YouTube Shorts and Google Flow. The Verge called it one of the most consequential Google I/O keynotes in years.
The DuckDuckGo surge is a real signal, but context matters: DuckDuckGo growing 37% still leaves it at a fraction of Google's search volume. What the data actually reveals is a latent demand for search-without-AI among a privacy-conscious segment — and that segment is now organizing itself around the Google changes. For marketers and SEO practitioners, this week's I/O aftermath should be a forcing function: your search traffic assumptions from even six months ago are stale. AI-generated answers in the main search box change click-through dynamics dramatically. If your analytics don't break down "AI Mode" impressions from traditional blue-link traffic, start building that visibility now before the numbers move further.
5. EU AI Act Gets a 16-Month Reprieve for High-Risk Systems as "AI Omnibus" Deal Lands
European regulators finalized what's being called the "AI Omnibus Agreement" this month, delivering significant deadline relief to companies implementing the EU AI Act — while also adding new prohibitions and simplifications. The most consequential change: obligations for Annex III high-risk AI systems (use-based) have been postponed from August 2, 2026 to December 2, 2027 — a 16-month extension. Obligations for Annex I high-risk AI systems (safety-critical sectors like medical devices and aviation) follow a separate two-tiered schedule.
"The agreement simplifies parts of the EU AI Act by: extending key compliance deadlines for high-risk AI systems; adding new prohibitions on AI-generated non-consensual intimate imagery and child sexual abuse material; and clarifying several definitional ambiguities that had created uncertainty for developers."
The AI Omnibus also adds explicit new prohibitions — including on AI-generated non-consensual intimate imagery (NCII) and AI-generated child sexual abuse material — while extending protections to whistleblowers surfacing safety violations. The agreement is expected to be formally adopted before August 2, 2026 when the original HRAIS compliance deadline was set to trigger. For companies that had been racing to meet the August deadline, the 16-month reprieve is operationally meaningful — but legal experts caution that the new December 2027 deadline is firm, and that using the extension to slow compliance programs would be strategically shortsighted given the complexity of what's required.
The 16-month extension is real breathing room, but it's a trap for companies that treat it as permission to pause. EU AI Act compliance for high-risk systems requires building data governance infrastructure, bias testing methodologies, incident reporting workflows, and third-party audit relationships — none of which can be stood up in the final months before a deadline. The smarter move: use the extra time to build compliance programs properly rather than reactively. Companies that treat this extension as a sprint extension will find themselves in the same scramble in late 2027. The new NCII and child safety prohibitions, by contrast, are effective immediately — if your AI products touch user-generated content, this needs legal review now.
6. Peter Diamandis on the "Organizational Singularity": AI Agents Will Restructure Companies Before Leadership Can Adapt
In his latest Moonshots podcast episode (recorded May 26), futurist and XPRIZE founder Peter Diamandis tackled what he calls the "organizational singularity" — the point at which AI agents, AI-native workflows, and recursive self-improvement restructure companies faster than traditional corporate hierarchy can adapt. The episode, titled "AI-Proof Your Company," is Diamandis's most direct statement yet about the pace at which agentic AI will transform the operating fabric of enterprises.
Diamandis argues that the companies best positioned for the organizational singularity share three traits: they have strong internal data moats (not just access to AI models), they have mapped their workflows at a process level (not just a departmental level), and they have executives who treat AI deployment as a core strategic competency rather than an IT initiative. His warning: traditional hierarchies aren't just slow to adapt — they actively resist the kinds of continuous-improvement loops that AI-native organizations run on.
"Kids will need qualities such as curiosity, purpose, and adaptability to succeed in the AI era. It won't matter what you studied — it will matter how fast you can learn and relearn."
— Peter Diamandis, via Business Insider, May 2026
Diamandis's framing is well-timed given this week's other news: Anthropic closing a $965B raise while simultaneously shipping a new flagship model illustrates exactly the kind of recursive capability improvement cycle Diamandis describes. The organizational challenge isn't keeping up with a single AI model — it's building the institutional muscle to absorb continuous capability step-changes without losing strategic coherence.
The "organizational singularity" framing is more practically useful than most AI strategy rhetoric, precisely because it points to the organizational bottleneck rather than the technology one. Most enterprises already have access to powerful AI tools. The constraint is how quickly their decision-making structures, incentive systems, and talent pipelines can incorporate AI outputs into real operations. Companies that treat AI transformation as a technology project will keep hitting this wall. Companies that treat it as an organizational design challenge — and bring in leadership that has actually restructured workflows around AI agents, not just experimented with chatbots — will pull ahead. The window for that repositioning is narrowing fast.
7. Massive Study: Generative AI Now Outperforms Average Humans on Creativity Tests
A significant new study published this week — comparing more than 100,000 people against today's most advanced AI systems — found that generative AI can now outperform the average human on certain standardized creativity tests. The research, highlighted by ScienceDaily, used established divergent thinking assessments (measuring the ability to generate novel, varied ideas across multiple domains) and found that leading AI models scored above average human performance across multiple creativity dimensions.
The study is careful not to claim AI is "more creative" than humans in a holistic or experiential sense — the creativity tests measure specific dimensions of ideation speed and breadth, not aesthetic judgment or emotional resonance. But the results land at a consequential moment: as AI is being deployed to assist with marketing copy, product design, and strategic brainstorming, the study provides empirical grounding for conversations about where human creative judgment adds genuine differentiated value versus where AI-assisted ideation can accelerate the process.
The findings echo observations from practitioners across industries who report that AI tools are increasingly useful in early-stage creative work — generating initial options, exploring edge cases, and stress-testing assumptions — while human judgment remains critical for selection, refinement, and final execution against strategic context that AI lacks access to.
The headline — "AI beats humans at creativity" — will be weaponized in both directions. Don't let either extreme define your strategy. The more precise read: AI now performs above average on measurable ideation tasks, which means it can serve as a genuine force multiplier for creative professionals who use it in the ideation phase. The human edge shifts upstream (toward strategic framing and intentionality) and downstream (toward judgment, taste, and emotional authenticity). The implication for enterprises: AI-augmented creative teams aren't just cheaper, they're faster and broader in their ideation. Businesses that haven't integrated AI into their creative workflows yet are now operating at a structural speed disadvantage relative to competitors who have.
Why All of This Matters Together
This week's stories aren't isolated developments — they form a coherent picture of an industry entering a new phase. Capital is no longer scarce for the leading AI labs; Anthropic's $965 billion raise effectively ends questions about whether it can compete at scale. The model layer is maturing: OpenAI is retiring transitional models and imposing real lifecycle discipline. Regulation is fragmenting between a hands-off federal government and aggressive state and EU-level action. And the user backlash against Google's AI-first Search signals that not everyone is rushing to embrace ambient AI — a reminder that adoption curves are never as clean as vendor roadmaps suggest. For enterprises, the strategic imperative hasn't changed: build internal AI competency fast, choose vendors with staying power, and treat compliance planning as a first-class operational capability. The organizations that do all three simultaneously are the ones that will be well-positioned when these technology curves steepen further.
Need help navigating AI for your business?
Our team turns these developments into actionable strategy — from model migration planning to AI governance frameworks and agentic workflow design.
Contact SEN-X →