When the CEOs of the three top AI labs sit in the same room before G7 leaders for the first time, when ChatGPT breaks the all-time speed record to 1 billion users, and when a new attack can hijack AI coding tools with 85% success — the AI industry is simultaneously living through three historic moments of governance, scale, and security.
G7 Summit: Three AI CEOs Face World Leaders Together
From June 15-17, 2026, the G7 summit in Évian-les-Bains, France, welcomed the CEOs of the world’s three leading frontier AI labs for the first time. According to the guest list published by the French presidential office, OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, and Google DeepMind CEO Demis Hassabis all confirmed their attendance.
This is the first time in history that the leaders of the three largest AI companies have appeared simultaneously before the heads of state of the world’s seven largest advanced economies. France, holding the rotating G7 presidency this year, has placed AI at the center of the summit agenda. Altman attended at the personal invitation of French President Emmanuel Macron — marking his first G7 appearance.
OpenAI’s Chief Global Affairs Officer Chris Lehane said the company expects tech firms to agree to a package of voluntary commitments at the summit, with youth safety as Altman’s personal priority. Lehane also cited frontier AI risks in the cyber and biological domains — a direct reference to the central issues in the Fable 5 export control dispute.
Adding to the drama, Amodei is attending the G7 summit while Anthropic is simultaneously in dispute with the US Commerce Department over the Fable 5 model suspension order. This creates an extraordinary situation: Anthropic’s CEO is at a G7 meeting on AI governance while his company’s flagship model remains disabled by a federal directive.
The three CEOs also recently co-signed a letter to Congress advocating for stricter regulations on synthetic DNA and AI-related biological threats — a rare moment of unity among competitors on a sensitive issue.
AI Pulse View: The G7’s decision to bring all three AI CEOs into the same room for the first time marks a watershed moment: AI governance has officially transitioned from “industry self-regulation” to “global political negotiation.” This is no longer a narrative about competition between tech companies — it’s about how nations will shape the rules of AI. Notably, the three CEOs arrive with divergent positions on sensitive issues: Amodei is caught in a federal dispute, Altman comes with Macron’s personal invitation, and Hassabis carries Google’s global compliance infrastructure. This “same room, different agendas” dynamic captures the core challenge of AI governance: when technological capability outpaces existing international frameworks, who writes the rules, who enforces them, and who watches the rule-makers? The G7 outcome may not change the status quo overnight, but it establishes a precedent — AI is no longer just a tech issue. It now sits alongside nuclear non-proliferation and climate change as a matter of global governance.
ChatGPT Hits 1 Billion Users: The “Not Winner-Takes-All” AI Market
According to estimates from market intelligence firm Sensor Tower, OpenAI’s ChatGPT crossed 1 billion global monthly active app users in May 2026 — becoming the fastest consumer app in history to reach that milestone, surpassing TikTok, Instagram, and YouTube at equivalent points in their growth trajectories.
This milestone carries enormous weight for OpenAI: the company has secretly filed for a September 2026 IPO at an $852 billion valuation, and the 1 billion user count is the single most important proof point for its revenue growth story.
But the more revealing data lies in the growth rates of competitors:
| AI Assistant | Monthly Active Users | YoY Growth |
|---|---|---|
| ChatGPT | 1 billion | Fastest consumer app to 1B in history |
| Meta AI | ~600M+ (estimated) | +973% |
| Claude | Not disclosed | +640% |
Claude’s 640% year-over-year growth, on an already substantial base, represents the kind of compounding trajectory that supports Anthropic’s $965 billion pre-IPO valuation. Meta AI’s 973% growth makes it the fastest-growing AI assistant globally — primarily driven by WhatsApp and Instagram integration, distributing the assistant to billions of users who never had to download a new app.
Meanwhile, CNBC reported that despite souring public sentiment toward AI, ChatGPT’s user growth remains strong. This reveals a significant trend: a “knowledge-action gap” exists between public concern about AI (privacy, safety, job displacement) and actual product usage — people express worry while simultaneously using AI products heavily in their daily lives.
AI Pulse View: One billion users is not just a number — it marks AI’s transition from “early adopters” to “mass infrastructure.” But the real industry insight is that the AI market is not consolidating toward a single winner. ChatGPT, Claude, and Meta AI are all growing on steep trajectories simultaneously, suggesting the AI assistant market will resemble the search engine market (Google, Bing, DuckDuckGo coexist) rather than the social network market (Facebook dominance). Meta AI’s 973% growth, powered by platform distribution, is a reminder to all AI companies: model capability matters, but distribution channels and ecosystem integration are equally decisive. For OpenAI’s IPO narrative, 1 billion users is the core story — but the conversion efficiency (paid conversion rate, ARPU) is what will ultimately determine whether the valuation holds.
Agentjacking: New Attack Hijacks AI Coding Agents at 85% Success Rate
Security firm Tenet Security disclosed a new attack class called “Agentjacking” in mid-June 2026. The attack exploits AI coding agents’ (Claude Code, Cursor, OpenAI Codex) trust in external error tracking services (Sentry) by injecting forged Sentry DSNs (Data Source Names), causing AI coding agents to automatically execute malicious code on developers’ machines.
Key characteristics of the attack chain:
- No phishing required: Attackers don’t need to trick developers into clicking malicious links
- No malware delivery: No software needs to be installed on the victim’s machine
- No server compromise: No prior server access is needed
- 85% exploitation success rate: In test environments, the attack almost always succeeds
- 2,388 organizations exposed: Security researchers have scanned and found nearly 2,400 organizations with affected AI coding tool configurations
The core mechanism exploits Sentry’s open-source nature — Sentry DSNs are publicly accessible, meaning anyone with a project ID can send forged error reports to that DSN. When AI coding agents (like Claude Code) are configured with a Sentry MCP server to automatically analyze and fix errors, attackers simply need to send forged error reports containing malicious instructions, and the AI coding agent will “faithfully” execute them.
Recommended mitigation: Remove Sentry MCP server configurations from Claude Code, Cursor, Codex, and other AI coding agents until Sentry implements authenticated event submission.
AI Pulse View: Agentjacking represents a fundamentally new paradigm of security threats in the AI era. Traditional cybersecurity attacks target system vulnerabilities or human error — Agentjacking targets the trust mechanism of AI agents themselves. When AI coding agents are designed to automatically trust input from external services like Sentry, they effectively expand the attack surface to any signal that can be forged. This is not just a Claude Code or Cursor problem — any AI agent integrating MCP servers or similar external tool-calling mechanisms faces analogous trust chain risks. The deeper insight: the more autonomous action an AI agent has (executing code, modifying files, deploying services), the greater the destructive potential when its trust chain is hijacked. Agentjacking’s 85% success rate is a stark reminder — until AI agent security frameworks mature, the design assumption of default trust in external inputs needs to be fundamentally re-examined.
Industry Implications: AI Enters an Era of “Global Governance + Super Scale + Security Unknowns”
The convergence of these three events sketches a portrait of the AI industry at a historical inflection point:
-
Governance: The G7 placing three AI CEOs in the same room for the first time elevates AI from a technology issue to a global governance issue. But the CEOs arrive with different national relationships, regulatory positions, and corporate interests — consensus will not come easily.
-
Market: The 1 billion user milestone proves AI products are spreading faster than any precedent, but the “not winner-takes-all” multipolar growth pattern means competition will be more intense and sustained.
-
Security: Agentjacking exposes the structural fragility of AI agent trust chains — when AI agents are given autonomous execution capabilities, traditional cybersecurity boundaries are being redefined.
AI Pulse View: Read together, these three events from mid-June 2026 reveal a deeper industry trend: AI is transitioning from the “technology experiment” phase to the “social infrastructure” phase. As infrastructure, it must endure the scrutiny of global governance (G7), support a super-scale user base (1 billion users), and confront the security threats that come with it (Agentjacking). This is not “growing pains” for the AI industry — it is the path that all world-changing technologies (the internet, mobile communications, search engines) have gone through. The difference is that AI’s adoption speed is faster than all of them, leaving less time for the industry to adapt. For AI practitioners and businesses, this means three things must be done simultaneously: embrace governance rather than avoid it, build sustainable business models within scale growth, and redesign security boundaries even as AI is given more autonomy.