The past few days have delivered a reminder that AI's pace hasn't slowed even as safety anxieties mount. Three storylines in particular capture where the industry stands heading into the back half of 2026.

Mistral swings for the fences with Large 4

Paris-based Mistral AI used this week to stake its claim as Europe's frontier-model champion. Mistral said it has started a public preview of its new flagship model, Mistral Large 4, internally nicknamed "le Chonk," a sparse mixture-of-experts model with 1 trillion total parameters, 49 billion of which are active at inference, and natively multimodal from the ground up. That architecture choice — routing each query through only a fraction of the total network — is what makes a trillion-parameter model commercially viable to serve at all.

Notably, Mistral wants to offer organisations a powerful model they can eventually run under their own control, with the model weights scheduled for release by the end of the month. The company is also leaning hard into enterprise use cases: for coding, Mistral reports a combined Coding Agent Index result of 49.8%, placing the model above DeepSeek V4 Pro and Qwen3.8 Max. It's a direct challenge to the assumption that only Chinese labs are pushing the open-weight frontier forward.

While Mistral charges ahead, OpenAI spent the week walking back a planned release. According to reporting on the company's internal safety process, OpenAI has withdrawn the planned October release of GPT-6.1 Astra after internal testing revealed the model fell short of the company's safety and alignment standards. In its place, the company shipped a cheaper sibling: OpenAI launched GPT-6.1 Sol at $2 per million input tokens and $10 per million output tokens, one-fifth of GPT-6 Astra's standard prices, saying it delivers near-Astra performance for agentic coding, computer use and professional work.

The episode arrives alongside broader unease about autonomous agents. At a New York City Council hearing this week, OpenAI, Anthropic, Meta and Google representatives stopped short of guaranteeing their AI agents will always obey safeguards. JPMorgan's chief executive added to the chorus of concern, warning that the cybersecurity risks posed by artificial intelligence increased dramatically after Anthropic introduced its powerful Mythos model.

Washington stands up a Super Intelligence Force

Government is scrambling to keep pace. President Donald Trump announced the formation of what he called a "Super Intelligence Force" to "ensure that America continues to lead the World" in AI. Trump said the task force will be led by Director of National Intelligence Jay Clayton, along with the Chairman of the FTC, the Under Secretary of Defense for Research and Engineering, and the Director of the Office of Personnel Management.

Taken together, the week's news underscores a widening gap between capability and control: models are getting bigger and cheaper even as the firms building them, and the governments overseeing them, openly admit they can't yet promise those systems will behave.

Sources

AI PolicyAI SafetyArtificial IntelligenceDaily AI UpdateMistral AIOpenAI