AI agents, export curbs and deepfake ads test tech's next phase

Qualcomm CEO Cristiano Amon described a near future in which AI agents supplant traditional apps, driving demand for more than 40 new device designs that include wearable cameras, jewelry, and always-on pins capable of observing their surroundings. These agents would handle complex, multi-step tasks such as retrieving banking details or booking travel by drawing context from the physical world rather than relying on screen-based interfaces. The shift, Amon argued, will produce broad experimentation with form factors that keep computing constantly available and conversationally accessible.

That hardware trajectory collides with sudden shifts in U.S. export policy. The Trump administration placed Anthropic’s most advanced models under export controls, a move that effectively requires case-by-case approval for certain foreign users. Canadian Prime Minister Mark Carney warned that over-reliance on U.S. models now carries political risk, prompting the European Union to accelerate its own data-center and chip initiatives. The precedent suggests that even voluntary regulatory frameworks can quickly harden into licensing systems when administrations change priorities, leaving allied governments uncertain whether American AI will remain a stable foundation for their own infrastructure.

The same generative capabilities powering Qualcomm’s envisioned agents are already reshaping political advertising. In Texas, an ad funded by a Trump-aligned group placed Democratic Senate candidate James Talarico in digitally fabricated scenes singing about transgender children. Similar tactics appeared in Kentucky’s Republican primary, where deepfakes depicted Representative Thomas Massie in fabricated personal situations. Campaigns on both sides have used the technology, yet disclosure remains voluntary and enforcement minimal. Democrats have signaled that any regained congressional majority would seek mandatory labeling, but the technical ease of production continues to outpace legislative responses.

Taken together, the developments illustrate a widening gap between technical possibility and institutional capacity. Device makers are racing to embed agents that can act across services and physical contexts, while regulators experiment with controls that risk fragmenting global access. At the same time, electoral norms that once rested on shared expectations of authenticity are eroding without clear replacement rules. The United States retains advantages in chip design and frontier model development, yet those advantages depend on predictable policy and credible democratic guardrails that current trends have not yet secured.