May 29th, 2026 | As AI IPOs loom, the Pope and Democrats call for oversight
- Justin Flechsig
- Jun 15
- 4 min read
The leading AI companies are now valued in the hundreds of billions, the Pope has written 42,000 words on why that should worry us, and healthcare cyber security executives are clambering to be included on Anthropic's Project Glasswing. Simultaneously, Progressive Democrats are converging on AI regulations, while the Trump administration warns states hands off.
This week: Mythos and the security access CIOs are lobbying to get, the AI regulation debate heats up, and a compliance study finds every major LLM fails EU law
Policy, Regulation, and Culture
On Regulating AI:
Illinois passes what may be the country's most stringent state AI oversight law: SB 315 would require frontier AI labs — including OpenAI, Anthropic, and Google — to have their safety practices verified by independent third-party auditors, going further than existing California and New York rules, which stop at guardrails and incident reporting. Governor Pritzker says he will sign it.
Progressive Democrats converge on anti-AI proposals: Five prominent progressives — Sanders, AOC, Warren, Khanna, and Congressional Progressive Caucus Chair Greg Casar — are advancing distinct but aligned measures including a data center moratorium, AI company taxes, and worker displacement protections, signaling AI is crystallizing into an organized intra-party faction rather than scattered criticism.
AI billionaires float redistribution to blunt backlash: Axios reports that AI leaders are floating redistribution frameworks — Bezos wants zero federal income tax for the bottom 50%, Altman favors universal basic compute, and OpenAI's foundation committed $250 million to worker mitigation — as California unions say they gathered over 1.5 million signatures for a billionaire wealth-tax ballot measure.
The Pope Sends a Memo to Silicon Valley
History rhymes. Pope Leo XIV released his first major encyclical, "Magnifica Humanitas," a 42,300-word document warning that AI risks deepening inequality, displacing workers into "forced inactivity," and making warfare more autonomous and psychologically distant. The timing was deliberate: he signed it on May 15, exactly 135 years after Leo XIII's Rerum Novarum — the Church's foundational text on labor in the Industrial Revolution — drawing a direct line between that moment and this one.
Rejection or Reckoning? Leo is careful to say technology is not "antagonistic to humanity," but the encyclical's sharpest language targets the concentration of AI development among a small number of companies and the failure of governments to impose meaningful oversight — arguing that "the invisible hand of the market" is insufficient and calling for robust legal frameworks, independent oversight, and clear accountability at every stage of AI decision-making. He presented the document alongside Christopher Olah, a co-founder of Anthropic.
The health system read. For clinicians and executives, the encyclical's most operationally relevant argument is accountability: who must "account for decisions, justify them, monitor them, and remedy any harm caused."
The Latest in AI
Frontier AI Advancements
Anthropic releases Opus 4.8 with multi-agent orchestration tool: The new model arrived 41 days after Opus 4.7, alongside a research preview of Dynamic Workflows, designed to coordinate tasks across hundreds of parallel subagents. Anthropic says the model is less likely to make unsupported claims; early tester, Bridgewater Associates, reported it flagged input and output issues other models missed.
Perplexity embeds its Computer product in Microsoft 365: Perplexity says its Computer tool is now available inside Excel, Word, PowerPoint, and Outlook as a side-panel assistant for drafting, modeling, and email.
AI Trends and Chatter
Anthropic overtakes OpenAI by valuation: Anthropic raised $65 billion in a Series H round, valuing the company at $965 billion — above OpenAI's most recent $730 billion valuation. Anthropic says run-rate revenue crossed $47 billion this month; neither figure has been independently verified.
Google Search AI dominates the news cycle:
Pichai describes a Google reorganized around AI: In a post-I/O interview, the CEO outlined structural changes including merged AI research arms, centralized infrastructure, and weekly AI product reviews, while acknowledging Search's new opinionated answers may be over-weighting Google's own synthesis.
AI search is keeping users inside Google: Ahrefs data cited by The Register puts the average clickthrough rate drop for top-ranked pages at 58% since AI Overviews launched, up from 34.5% eight months earlier, while Google Search VP Liz Reid says AI Mode queries have more than doubled every quarter.
Google closes developer tools and embeds ads in AI answers: The Register's commentary flags Google's deprecation of the open-source Gemini CLI in favor of a closed-source enterprise product, alongside plans to insert ads directly into AI Mode responses.
Hospitals and Care Delivery
Health system CIOs are urging Anthropic to include healthcare organizations in Project Glasswing, a program giving critical infrastructure firms private access to the Mythos AI model to identify and fix cybersecurity flaws.
The VA has deployed ambient AI to reduce documentation burden for clinicians serving 9 million veterans, with recent staffing cuts adding pressure to a system already strained by heavy charting requirements.
Mayo Clinic received over 450 AI agent ideas from its nurses, advancing 33 to a build phase with Microsoft in a program designed to let nurses design tools for their own clinical workflows.
Research, Evidence, and Benchmarks
Nonprofit Aithos tested every major LLM against EU law using its LARA tool and found all of them violated GDPR and AI Act requirements across scenarios including user manipulation and covert data monitoring, with the top scorer reaching only 54% compliance.
A study in the Journal of the American Podiatric Medical Association compared ChatGPT-5 and DeepSeek V3 on patient education quality for 35 foot and ankle disorders, finding ChatGPT-5 scored higher on reliability and comprehensiveness while DeepSeek V3 produced simpler, more readable text.