AI News Roundup — June 16, 2026
The three most powerful AI CEOs — Altman, Amodei, and Hassabis — appeared before G7 world leaders together for the first time ever, with Amodei doing so while his flagship model remains disabled by federal order. SpaceX exercised its option to acquire Cursor for $60 billion in an all-stock deal, the largest AI developer tools acquisition in history. A new 'Agentjacking' attack is actively exploiting Claude Code, Cursor, and Codex with an 85% success rate against 2,388 exposed organizations. And Google shipped Android 17 packed with Gemini AI features.
All Three AI CEOs Appear Before G7 Leaders Together for the First Time
Sam Altman (OpenAI), Dario Amodei (Anthropic), and Demis Hassabis (Google DeepMind) all attended the G7 summit in Évian-les-Bains, France — confirmed via the French presidential office guest list. It marks the first time the CEOs of the three most powerful AI companies have appeared simultaneously before the leaders of the world's seven largest advanced economies. France holds the G7 presidency and placed AI prominently on the agenda: frontier AI risks in cyber and biological domains, youth safety, and voluntary industry commitments. Altman attended at the personal invitation of President Macron; OpenAI's global affairs chief Chris Lehane said tech companies were expected to leave having agreed to a package of voluntary commitments. The summit created an extraordinary juxtaposition for Amodei: simultaneously at a G7 meeting on AI safety governance while his company's Fable 5 and Mythos 5 models remain disabled by a US Commerce Department directive — the very cyber and bio risks on the G7 agenda. All three CEOs also recently co-signed a letter to Congress calling for stricter regulations on synthetic DNA and AI-related biological threats, a rare moment of cross-competitor alignment.
SpaceX Acquires Cursor Maker Anysphere for $60 Billion — Largest AI Dev Tools Deal in History
SpaceX exercised its option to acquire Anysphere, the company behind the AI coding tool Cursor, for $60 billion in an all-stock deal expected to close in Q3 2026. The acquisition is the largest in AI developer tooling history. SpaceX had secured the option in April 2026 — at the time offering either $60 billion to acquire or $10 billion for a partnership — and exercised the full acquisition option four days after SPCX's record-breaking IPO debut. Cursor had approximately 4 million daily active developers at the time of the deal. The move gives SpaceX a direct competitive position in the AI coding agent market against Claude Code (Anthropic), Codex (OpenAI), and GitHub Copilot (Microsoft) — and transforms Grok Build, xAI's terminal coding agent launched in early beta last week, into one component of a broader coding platform. SPCX's post-IPO stock surge made an all-stock deal structure far more attractive than at any earlier point.
Reuters · New York Times · Yahoo Finance
Agentjacking — New Attack Exploits Claude Code, Cursor, and Codex at 85% Success Rate
A new attack vector called "Agentjacking" is actively targeting AI coding agents including Claude Code, Cursor, and Codex, with an 85% exploitation rate in testing and 2,388 organizations confirmed as exposed. The attack works by injecting malicious instructions into fake Sentry error messages — Sentry is the error monitoring tool used by the majority of professional development teams. When an AI coding agent reads what appears to be a legitimate Sentry error report, it follows attacker-controlled instructions embedded in the fake message, executing arbitrary commands inside the developer's codebase and CI/CD pipeline. The attack is a sophisticated form of prompt injection — one of the most fundamental security vulnerabilities in agentic AI systems. Recommended immediate actions: audit Sentry DSN and webhook configurations for unauthorized modifications; review recent AI agent activity logs for unexpected commands or file modifications; restrict AI agent permissions to limit directory and command access; treat all external data sources — error logs, issue trackers, PR comments — as untrusted input in agentic contexts. Agentjacking joins a self-replicating worm that hit 73 Microsoft GitHub repositories through AI coding tools earlier this month, forming a concerning pattern as AI agents gain deeper access to production systems.
Google Ships Android 17 and Wear OS 7 — Gemini Omni, Lyria 3, and AudioLM Across Pixel Devices
Google released the final version of Android 17 on Tuesday, accompanied by Wear OS 7 and a Pixel Drop packing new AI capabilities. Headlining features: Gemini Omni now lets users edit videos in conversation; Lyria 3 lets users create music tracks from text prompts and images in the Gemini app; Pixel 10a gets speech-to-speech translation tools powered by AudioLM. On the OS side, Android 17's "bubble bar" lets users organize recent apps as movable bubbles along the bottom of the screen. A new simultaneous selfie-and-screen recording feature targets content creators. Parental controls were expanded to include screen time limits and content filtering set with a PIN without requiring a Google account. Wear OS gains Gemini Intelligence features this summer including personalized widget creation from text descriptions and Personal Intelligence — connecting Google app history with Gemini. Quick Share becomes compatible with Apple AirDrop. The launch contrasts sharply with Apple's approach: while Google ships natively integrated AI models to Android 17, Apple's iOS 27 AI features aren't expected to fully rollout until September.
The AI Assistant Market Is Not Winner-Takes-All — ChatGPT at 1B, Meta AI at 973% Growth
The latest AI assistant data confirms what was previously uncertain: the market is not consolidating around a single winner but expanding with multiple platforms growing simultaneously at extraordinary rates. ChatGPT reached 1 billion monthly active users — the fastest consumer app in history to achieve that milestone, surpassing TikTok, Instagram, and YouTube at equivalent growth stages. Claude grew 640% year-over-year. Meta AI grew 973% year-over-year — the fastest-growing AI assistant globally, driven primarily by passive distribution through WhatsApp and Instagram rather than active downloads. The multi-platform growth pattern has significant implications for the upcoming AI IPO wave: both Anthropic and OpenAI can tell credible revenue growth stories, and neither needs to argue for winner-take-all dynamics to justify its valuation.
Critical Copilot Vulnerability Let Hackers Steal 2FA Codes — AI-Integrated Tools Expand Attack Surface
A critical security vulnerability in Microsoft Copilot allowed attackers to steal two-factor authentication codes from users, Ars Technica reported. The flaw underscores a pattern developing alongside the broader AI coding tool ecosystem: as AI assistants gain deeper integration into developer environments, authentication systems, and enterprise workflows, each new integration point creates potential new attack vectors. The Copilot vulnerability, the Agentjacking exploit targeting Claude Code and Cursor, and the earlier self-replicating worm in Microsoft's GitHub repositories all landed within the same two-week window — suggesting the security community has not yet caught up with the attack surface created by widespread AI agent deployment. The timing is particularly notable given that SpaceX's $60 billion Cursor acquisition and Anthropic's $965 billion IPO valuation are both predicated on the continued growth of AI coding tool adoption.
Alibaba Unveils AI Models for Robots as China's Focus Shifts to Agents
Alibaba released new AI models specifically designed for robotics applications, signaling a strategic shift from large language model competition to agentic and embodied AI as China's next frontier. The models are designed to give robots better grounded reasoning, physical scene understanding, and task-level planning — capabilities required for robots that need to operate in unstructured real-world environments rather than controlled factory floors. The announcement follows EngineAI's Hong Kong IPO filing and Unitree's $7 billion listing plans, and fits within China's broader pattern: having established manufacturing dominance in humanoid robot hardware, Chinese labs are now racing to close the software and AI capability gap that currently limits humanoid robots to structured industrial use cases.