This week in AI and tech delivered three defining stories that reshaped the industry landscape — a geopolitical crackdown on frontier models, the quiet emergence of GPT-5.6 in the wild, and a surprising market share milestone: ChatGPT below 50% for the first time since launch.
Fable 5: Three days that shook the industry
Anthropic launched Claude Fable 5 on June 9 — the company's first publicly available "Mythos-class" model, boasting a 1-million-token context window and record-breaking performance on complex analytics and software engineering benchmarks. Three days later, the US Commerce Department ordered a global access shutdown over a reported jailbreak vulnerability.
More than 100 cybersecurity leaders signed an open letter calling the ban disproportionate, warning it would harm American competitiveness. Anthropic CEO Dario Amodei met with Trump administration officials, but no resolution has emerged. The Economist ran a cover story titled "America's AI Power Grab," framing the episode as a watershed moment for geopolitical control of frontier AI.
Chinese labs, meanwhile, took note. Zhipu AI's GLM-5.2 beat GPT-5.5 on the FrontierSWE benchmark by a narrow margin — within one point of the now-disabled Fable 5.
GPT-5.6 already live for Pro users
Reports from ChatGPT Pro subscribers suggest GPT-5.6 is already active in the tier, delivering noticeably faster and more capable performance — completing end-to-end coding tasks like building a browser game in a fraction of the time. OpenAI's Chief Scientist described it as a "meaningful improvement." A full public launch is expected by late June.
The timing looks strategic: OpenAI is aiming to regain momentum just as the competitive landscape fragments.
ChatGPT below 50% for the first time
According to Sensor Tower, ChatGPT's global AI assistant market share has dropped to 46.4% — the first time below the 50% threshold since launch. Google's Gemini holds 27.7%, and Claude sits at 10.3%. Despite the relative decline, absolute monthly users remain at 1.1 billion, but the shift underscores the rise of competitor ecosystems, especially in agent-based workflows.
Karpathy warns of a widening understanding gap
Andrej Karpathy — who recently moved from OpenAI to Anthropic — posted a sharp critique on X, warning of a "growing gap in understanding of AI capability." He argued that many people still judge modern AI based on outdated free-tier ChatGPT experiences, while models have improved dramatically. He described the phenomenon among developers as "AI Psychosis," with the general public likely next.
DeepSeek closes $7.4B round at $50B+ valuation
Chinese AI startup DeepSeek completed its first external funding round, raising over $7.4 billion at a valuation exceeding $50 billion, backed by Tencent, CATL, and state-linked funds. The round makes DeepSeek China's most valuable AI startup and one of the highest-valued private AI companies globally.
Agentjacking: A new security threat for AI agents
Security researchers disclosed a novel attack dubbed "Agentjacking," which exploits Sentry error-tracking infrastructure to inject malicious code into AI coding agents such as Claude Code and Cursor. The attack affected 2,388 organizations with an 85% exploitation rate, underscoring the growing risks of autonomous agent workflows without human oversight.
The bottom line
The weekend of June 21 paints a picture of an industry moving in three conflicting directions at once: a breakneck model arms race (GPT-5.6 incoming), a geopolitical regulatory clampdown that can kill a frontier model overnight, and a market that's rapidly fragmenting — ChatGPT is no longer alone at the top. Anyone who watched Fable 5 go dark after three days got a vivid reminder that frontier technology and international politics are now inseparable.
