Learn how Microsoft 365 Copilot Business delivers ROI

Good morning. It's Monday, March 30, and we're covering Anthropic’s massive IPO ambitions, Google’s surging internal AI agents, growing concerns around AI deception, and more.

If you enjoy this email, share it with a friend. First time reading? Sign up here to keep up with the future of tech.

Anthropic “Mythos” Model Leak Confirmed: Security researchers exposed ~3,000 internal files describing Claude Mythos (“Capybara”), a model Anthropic told Fortune is real and a “step change,” raising cybersecurity risk concerns.

AI Chatbots Show Rising Deception: UK-backed study finds nearly 700 real-world cases, with misbehavior up fivefold since October 2025, including rule-breaking, deception, and unauthorized data deletion.

Judge Blocks US Action Against Anthropic: Federal court orders reversal of “supply-chain risk” label and agency bans in March 2026 dispute, siding with company’s claims of unlawful government overreach.

Apple Offers Bonuses to Retain Talent (Paywall): Company gives iPhone designers six-figure bonuses to curb exits to AI firms like OpenAI, which are increasingly building competing hardware devices.

Alphabet Seen Gaining 40% on AI Strength: Wells Fargo lifts price target to $397, citing Gemini adoption, cloud growth, and TPU deals driving high-margin revenue expansion through 2027.

Microsoft Faces Worst Quarter Since 2008: Shares down 25% in Q1 2026 as heavy AI spending and rising competition from OpenAI and Anthropic weigh on growth and margins.

SK hynix Eyes $14B US IPO: Chipmaker files for 2026 listing to fund AI memory expansion amid “RAMmageddon” shortages and surging demand for high-bandwidth memory used in AI systems.

Matt unpacks ARC AGI benchmarks, highlighting AI’s struggles and the shift to the interactive ARC AGI 3 test.

Anthropic executives are discussing an initial public offering as early as the fourth quarter of 2026, with bankers estimating the deal could raise more than $60 billion. The AI firm, last valued at $380 billion, has seen rapid growth—annualized revenue doubled to $19 billion in early 2026—driven largely by enterprise and developer tools.

The move would position Anthropic ahead of rival OpenAI in public markets, as investors reportedly favor its business model and shorter path to profitability. Still, timing remains uncertain and could shift with market conditions, including inflation concerns tied to geopolitical tensions. The IPO would mark a major milestone in the AI boom, testing investor appetite for high-growth, capital-intensive AI firms. → Read the full article here. (Paywall)

Google employees are rapidly adopting an internal AI tool called “Agent Smith,” which automates coding and other workflows, prompting the company to restrict access due to high demand. Launched in early 2026, the tool builds on Google’s “Antigravity” platform and can autonomously execute tasks in the background, pulling from internal systems and responding via chat or mobile.

Executives, including cofounder Sergey Brin, are signaling that AI agents will be central to the company’s strategy this year. The surge in usage comes as Sundar Pichai pushes employees across roles to adopt AI tools, with usage increasingly tied to performance expectations. The tool underscores how AI agents are shifting from experimental features to core workplace infrastructure. → Read the full article here.

MIT professor Dimitris Bertsimas used his 2026 Killian Lecture to showcase how operations research and AI are improving sectors like healthcare, logistics, and education. His work on “robust optimization” has enhanced systems from Panama Canal shipping flows to school bus routing, while newer AI-driven tools are helping hospitals reduce patient stays and expand capacity.

In collaboration with Hartford HealthCare, his models cut average hospital stays from 5.38 to 4.93 days, enabling over 5,000 additional annual patients. Bertsimas also emphasized AI’s role in scaling education, with tools for summarization and translation supporting global online learning. The talk reflects a broader shift toward combining mathematical optimization with AI to deliver measurable, real-world gains. → Read the full article here.

A June 2024 essay by Leopold Aschenbrenner argues that rapid advances in compute and algorithms could make artificial general intelligence (AGI) plausible by 2027, with superintelligence soon after. The piece outlines an escalating “AI arms race,” with tech firms and governments expected to invest trillions in data centers, chips, and power infrastructure.

It warns that AGI could trigger an “intelligence explosion,” compressing decades of research into years, while raising major risks around security, alignment, and geopolitical conflict—particularly with China. The author frames the moment as akin to a large-scale industrial and national security mobilization. Much of the timeline and impact remain speculative. → Read the full essay here.

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team