DeepSeek Fallout, EU AI Act Enforcement, and DOGE's Cybersecurity Gutting — The Policy Earthquake Reshaping AI in 2026

Three converging policy shocks are redrawing the AI landscape in early 2026. Congressional momentum to ban DeepSeek from US government devices is accelerating, framing the open-source Chinese model as a national security threat comparable to TikTok. Simultaneously, the EU AI Act's first enforcement wave hit in February 2026, banning specific prohibited AI practices across member states and forcing enterprise compliance decisions at scale. Most critically, the DOGE efficiency operation's deep access to federal systems — including CISA — has created a paradox where the US is simultaneously hardening against Chinese AI while hollowing out its own cyber defense posture. The timing could not be worse.

Impact Score

9/10

Key Stories

POLICY

Congress Moves to Ban DeepSeek on Government Devices

Multiple congressional offices + Reuters — Bipartisan legislation treating DeepSeek as a national security threat — the TikTok playbook applied to AI. If passed, sets precedent for AI model origin restrictions. China-US AI decoupling accelerates.

POLICY

EU AI Act Prohibited Practices Enforcement Begins February 2026

European Commission — First hard compliance deadline under the AI Act. Banned: social scoring, real-time biometric surveillance in public, subliminal manipulation systems. US companies with EU operations now face real enforcement exposure — not just future risk.

CYBER

DOGE Access to CISA Systems Raises Insider Threat Concerns

Wired, Washington Post, CISA watchdog reports — DOGE team given broad access to federal IT systems including CISA infrastructure. Security researchers flagging this as an unprecedented insider threat surface. CISA workforce cuts compounding the issue — fewer defenders, wider attack surface.

OpenAI Operator and Anthropic Computer Use Drive Agentic AI Commercial Push

OpenAI, Anthropic product releases — AI agents that autonomously browse, click, and execute real-world tasks are entering broad commercial deployment. Attack surface for AI-specific exploits (prompt injection, agent hijacking) expanding rapidly — security tooling is not keeping pace.

POLICY

Trump AI Executive Order Rolls Back Biden Safety Framework

White House Executive Orders — Biden's AI safety EO revoked. NIST AI Risk Management Framework loses its mandate backstop. The US federal AI safety apparatus is now voluntary-only. This is the divergence point from EU regulation — and it will define the next 4 years of AI development.

Analyst Take

The 2026 AI story is not about model benchmarks — it's about control. Three actors are fighting over who governs AI and how: the US executive branch (deregulating domestically, weaponizing export controls externally), the EU (becoming the de facto global regulatory standard through market pressure), and Congress (reactive, using the national security frame to catch up). DeepSeek is the flashpoint because it broke the assumption that frontier AI required Western infrastructure. The DOGE-CISA situation is the underreported crisis — you cannot simultaneously gut your cyber defense agency and escalate geopolitical AI tensions without accepting serious exposure. Watch for the first major federal breach or AI-assisted state intrusion to land during this staffing trough. That's when the political calculus shifts fast. Dong should be tracking: EU AI Act enforcement actions (first penalties will set the fine baseline), any legislative text on the DeepSeek ban (defines what "Chinese AI" means legally going forward), and whether any of the gutted CISA roles get quietly rehired or backfilled via contractors.