Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.
Google's Threat Intelligence Group disclosed that it detected hackers using an AI model to discover and weaponize a zero-day vulnerability — a first. The exploit, a Python script bypassing two-factor authentication on a popular web admin tool, bore unmistakable LLM hallmarks including educational docstrings and a hallucinated CVSS score. Google says its proactive discovery likely prevented a "mass exploitation event."
Announced at Code with Claude, "dreaming" is a scheduled process that reviews an agent's past sessions, prunes stale notes, merges duplicates, and surfaces cross-session patterns no single run could see. Legal AI company Harvey reports 6x task completion improvement; Wisedocs cut document review time by 50%. Available now in research preview.
Anthropic's unreleased frontier model Claude Mythos can find and exploit zero-day vulnerabilities faster than expert pentesters, identifying thousands of critical flaws across every major OS and browser. Rather than release it broadly, Anthropic formed Project Glasswing — a coalition including AWS, Apple, Microsoft, Google, and CrowdStrike — giving roughly 40 organizations early access so defenders can patch before similar capabilities proliferate.
Musk officially killed xAI as a standalone company, absorbing it into SpaceX under the "SpaceXAI" brand. All 11 original co-founders had departed by late March. Meanwhile, Anthropic secured access to xAI's Colossus 1 data center — over 220,000 Nvidia GPUs and 300+ megawatts — to handle its 80x year-over-year usage growth. Musk called Anthropic "evil" three months ago; now he's their landlord for $4 billion.
Amazon Bedrock's new AgentCore Payments system, built with Coinbase and Stripe, lets AI agents autonomously pay for APIs, data feeds, paywalled content, and other agents via stablecoin micropayments. Developers set session-level spending limits; agents handle the rest. Warner Bros. Discovery is among the early testers. Future versions will expand to hotel bookings and merchant payments.
The Washington Post reports a policy battle inside the administration: intelligence agencies want pre-deployment vetting authority over frontier AI models, akin to FDA drug approval. The push was catalyzed by Anthropic's Mythos capabilities and the Google zero-day disclosure. CAISI has already signed agreements with Google DeepMind, Microsoft, and xAI to evaluate models before public release. Meanwhile, the Pentagon flagged Anthropic as a national security concern after the company refused to allow Claude for mass surveillance or autonomous weapons.
IBM's Institute for Business Value surveyed 2,000 global CEOs across 33 countries and found 76% have appointed a Chief AI Officer, up from just 26% in 2025 — a tripling in one year. Most CAIOs now report directly to the CEO or board. However, only 10% of organizations report growth-driven AI results; 53% remain in the piloting phase.
Google discontinued its experimental Mariner browser agent on May 4, folding the technology into a unified Gemini Agent. Meta is developing an agent called "Hatch" alongside an agentic shopping tool for Instagram, with internal testing set for end of June. Both companies are chasing Anthropic and OpenAI, who have pulled further ahead on agentic capabilities.
Alphabet shares have rallied 160% in one year, driven by the market's recognition that Google owns more layers of the AI stack than any competitor — custom chips (TPUs), foundation models (Gemini), cloud infrastructure, and distribution (Search, Android, YouTube). Analysts now view Google as one of two companies best positioned for long-term AI dominance.
Snap and AI search company Perplexity have quietly ended their planned $400 million AI partnership before it reached broad rollout. Details on the breakdown remain thin, but the collapse signals that not every big AI deal is a sure thing — even well-funded ones can fall apart when product-market fit doesn't materialize.
