Welcome to P3 Media’s AI Commerce Brief, your daily update on the AI and commerce stories shaping how companies build, sell, and grow. It’s Friday, June 26, 2026. Let’s get into it.
The top story today is about who gets access to frontier AI.
Axios reports that the Trump administration has asked OpenAI to limit the initial release of GPT-5.6 to a small group of government-approved partners, citing security concerns. The Verge reports that access during the preview would be approved case by case.
The reported OpenAI limits come shortly after a much sharper intervention involving Anthropic. On June 12, Anthropic said the US government issued an export-control directive requiring it to suspend access to Fable 5 and Mythos 5 by foreign nationals, whether inside or outside the United States. Anthropic said the practical result was that it had to disable those models for all customers in order to comply.
P3 Media's proprietary testing has shown Fable to be roughly three times more productive per token than Opus 4.8, largely because it uses subagents more efficiently to break down complex tasks.
The data points to the power of politicizing frontier model access. Businesses allowed to use government-restricted models will have a pronounced edge over competitors. Meanwhile, all teams building around frontier agents may need contingency plans across model providers, especially for workflows that depend on long-running, multi-agent execution.
OpenAI also published new research on how Codex is changing work.
The company says users are shifting from short chatbot-style interactions toward longer, delegated agent tasks. In sampled individual-user data, OpenAI says more than 80 percent of users had made at least one Codex request estimated to represent more than 30 minutes of human work by May 2026. More than 70 percent had made one estimated above one hour.
The data confirms that agentic work is becoming longer, more complex, and more cross-functional, even for non-developers, which OpenAI says are among the fastest-growing Codex user groups.
The implication is that as agents take on more complex tasks, agent orchestration is quickly becoming an essential workforce skill across functions.
Now today's Commerce Pulse.
Google is pushing Demand Gen deeper into paid-social territory.
Google says Demand Gen will add broader video resizing options, Gemini-powered creative recommendations, and Web-to-App Acquisition Measurement. The company offers Demand Gen ads across YouTube, Shorts, Discover, and Gmail, and says the inventory is designed for visually rich, multi-format campaigns.
By attempting to make YouTube and its surrounding surfaces behave more like performance-social channels, Google is putting Demand Gen directly into budget competition with Meta, TikTok, and other visual discovery inventory. The company has recently emphasized measurement tools that let advertisers compare Demand Gen performance against paid social, perhaps in a bid to win brands ready to spend on new-customer acquisition strategies beyond search.
Google's announcement comes as some companies raise concerns about AI ad automation brand controls.
Business Insider reports that REI blamed a Meta AI personalization tool for altering a vendor-supplied image in some Instagram ads. REI told Business Insider it had been auto-enrolled in the tool and later unenrolled.
AI creative systems can generate more variations and move faster, but there is currently no mechanism for maintaining product accuracy and brand standards across AI-generated creatives.
Global model watch:
Axios reports that China’s open-weight GLM-5.2 is raising concern among security researchers because models that can be downloaded and modified are harder for providers to monitor or restrict. According to Axios, researchers see the model as capable enough to matter for cyber workflows.
As the US government moves to restrict model access at home, contrasting access and control policies in China may create a first-mover advantage for companies with the capacity to adopt and refine overseas tech.
What to watch next: whether the reported OpenAI limited rollout becomes a template for frontier releases, whether agent workflows start changing operating models across non-coding functions, and whether Google can make Demand Gen a credible alternative to paid social budgets.
That’s your AI Commerce Brief for today. Thanks for listening.
