# Morning AI TL;DR — 2026-06-15

Fresh, high-signal AI news in the last 24–48 hours is relatively thin, but these are the most relevant updates for AI/dev/Salesforce ecosystem builders:

- **US reportedly restricted Anthropic model access after Mythos security concerns** — Al Jazeera/Reuters reported a US order requiring Anthropic to block foreign access to newly released high-end software, while The Verge reported concerns China may have accessed Mythos; this matters because frontier-model export controls could affect API availability, enterprise procurement, and multinational developer workflows. [Al Jazeera/Reuters](https://www.aljazeera.com/news/2026/6/13/us-orders-anthropic-to-disable-ai-models-for-all-foreign-nationals), [The Verge](https://www.theverge.com/ai-artificial-intelligence/949644/china-white-house-anthropic-mythos)
- **EU warns US AI export controls should not be discriminatory** — Euronews says the European Commission is scrutinizing implications for EU users, a sign that frontier-model access is becoming a trade-policy issue rather than just a vendor-policy issue. [Euronews](https://www.euronews.com/my-europe/2026/06/14/us-export-controls-on-anthropic-should-not-be-discriminatory-eu-commission-warns)
- **Z.ai’s ZCode 3.0.1 shipped with agentic coding optimizations** — The June 14 changelog says ZCode was optimized for GLM long-horizon reasoning, tool calls, and large-scale engineering execution chains, worth tracking as coding agents compete on real task completion rather than chat benchmarks. [Z.ai changelog](https://zcode.z.ai/en/changelog)
- **GLM 5.2 coverage points to 1M-context, coding-first open-weight ambitions** — Multiple developer-focused writeups report Z.ai’s June 13 GLM 5.2 launch with 1M context and coding-plan availability, with open weights/API expected next; if confirmed by official release notes, this is relevant for long-context repo and agent workflows. [Codersera](https://codersera.com/blog/glm-5-2-release-1m-context-coding-2026/), [BuildFastWithAI](https://www.buildfastwithai.com/blogs/glm-5-2-review-2026)
- **Databricks open-sourced Omnigent, a cross-agent meta-harness** — MarkTechPost reports an Apache-2.0 project for composing, governing, and sharing agents across Claude Code, Codex, and Pi; the strategic angle is standardizing multi-agent governance and reuse across enterprise stacks. [MarkTechPost](https://www.marktechpost.com/2026/06/13/databricks-open-sources-omnigent-a-meta-harness-that-composes-governs-and-shares-ai-agents-across-claude-code-codex-and-pi/)
- **Salesforce Agentforce adoption story is shifting from launch hype to revenue and governance** — Recent coverage highlights Agentforce ARR growth and near-term multi-agent/A2A orchestration plus voice testing, useful signals for builders deciding where to invest in Salesforce-native agent patterns. [Motley Fool](https://www.fool.com/investing/2026/06/11/salesforces-ai-business-is-growing-more-than-200-b/), [AMC Technology](https://www.amctechnology.com/resources/blog/tdx-2026-salesforce-is-serious-about-ai-trust-what-we-heard-on-the-ground)

**Worth watching:** Frontier AI may split into regional access tiers faster than expected; builders should design agent stacks with model-provider abstraction, audit trails, and graceful fallback paths.