GPT-5.6 on the Doorstep: OpenAI's Next Big Model Expected This Week
OpenAI appears ready to unveil GPT-5.6, its latest flagship model, with early testers in ChatGPT Pro already reporting faster and more capable responses. OpenAI Chief Scientist Jakub Pachocki described the model as a "meaningful improvement" over GPT-5.5, and Polymarket traders assign an 83% probability to a launch between June 22 and June 28.
Developer testing suggests GPT-5.6 supports approximately 1.5 million tokens of context — a 43% increase over GPT-5.5's documented 1 million token limit. Token usage per typical task reportedly drops an additional 10–15%, building on the efficiency gains already delivered by GPT-5.5. The release is expected to include standard, Mini, and Pro variants.
The timing is notable: OpenAI is pushing forward even as it navigates its acquisition of Astral, the startup behind the popular Python tools uv and ruff, which will be integrated into Codex — OpenAI's AI coding platform that now boasts over 2 million weekly active users.
ChatGPT Falls Below 50% Market Share for the First Time
In a milestone that underscores the rapidly shifting AI landscape, ChatGPT's share of the global AI assistant market has dropped to 46.4% as of May 2026, according to Sensor Tower data — down from 52.8% in December 2025 and 65.3% just a year before that.
Google Gemini has surged to 27.7% of the market (up from 18.2% in December 2024), while Claude has seen the most dramatic growth, jumping from 3% in December 2025 to 10.3% by May 2026. In absolute numbers, Claude's monthly active users quadrupled from 60 million to 245 million in just five months.
Despite the percentage decline, ChatGPT still leads in total users with 1.11 billion monthly actives. But the trend is clear: users are increasingly willing to switch between AI assistants, and specific controversies — such as OpenAI's Department of Defense partnership in February — have accelerated the shift.
'Agentjacking' Attack Exposes 2,388 Organizations Using AI Coding Agents
Security firm Tenet Security has disclosed a dangerous new attack vector called "agentjacking" that exploits AI coding agents through manipulated Sentry error reports. The attack works by posting fake error reports to Sentry's publicly accessible endpoints, which AI coding agents then retrieve and unwittingly execute as malicious code.
In controlled tests, Tenet successfully hijacked Claude Code, Cursor, and Codex with an 85% success rate. The researchers identified 2,388 exposed organizations, ranging from a $250 billion enterprise to solo developers. What makes agentjacking particularly insidious is what Tenet calls the "Authorized Intent Chain" — every step in the attack appears legitimate because the developer authorized the agent, the agent authorized the MCP connection, and the MCP connection returns data from a trusted service.
Sentry acknowledged the vulnerability but declined to implement a structural fix, describing the issue as "technically not defensible" at the platform level. Their only remediation has been a content filter targeting a specific payload string from the proof of concept — a band-aid that security experts say offers minimal protection.
Fable 5 Export Ban Faces Growing Opposition
The controversy surrounding the US export ban on Anthropic's Fable 5 and Mythos 5 models intensified this week. Over 100 cybersecurity leaders signed an open letter demanding reversal of the ban, calling the jailbreak vulnerability justification "disproportionate." Anthropic CEO Dario Amodei met with Trump administration officials but reached no resolution, with the White House reportedly demanding Anthropic achieve "zero jailbreaks" before any relaunch — a standard security experts consider technically impossible.
The ban's origins trace back to SK Telecom, South Korea's largest carrier and a $100 million Anthropic investor, which the White House identified as a security risk with Mythos access. After Anthropic revoked SK Telecom's access, Amazon researchers reported separate Fable 5 vulnerabilities, leading to the sweeping export control order that blocked all foreign national access.
At Anthropic's new Seoul office opening on June 18, Managing Director of International Chris Ciauri suggested the models would return "within days," though no official restoration date has been confirmed. Meanwhile, today marks the close of the Fable 5 free-trial window for paid Claude subscribers — a trial that ran just 4 usable days out of its planned 13 before the ban took effect.
Google Launches First Smart Speaker in Six Years with Gemini Built In
Google officially unveiled the Google Home Speaker, its first new smart speaker since the Nest Audio in 2020, with Gemini AI built in from the ground up. Priced at $99 and shipping June 25, the device represents Google's answer to Amazon's recently redesigned Echo (powered by Amazon Nova) and Apple's HomePod running Siri AI with Gemini integration.
The speaker features Gemini for Home, enabling natural multi-step conversations and intuitive smart home control without the rigid command structures of previous voice assistants. Premium features through Google Home Premium include Gemini Live for conversational interactions, Camera History Search for Nest cameras, and Home Briefs that summarize activity around your house. Available in four colors — Hazel, Porcelain, Jade, and Berry — with an underglow light ring that indicates listening and processing states.