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

Qualcomm chief executive Cristiano Amon outlined an aggressive push into artificial intelligence hardware last week, revealing that the chipmaker is developing more than 40 new device designs built around AI agents that could eventually supplant traditional smartphone apps. Speaking on CNBC's The Tech Download podcast, Amon described wearables such as camera-equipped earbuds, smart pins, jewelry and watches that would remain in constant contact with users, providing real-time context and conversational access to AI systems capable of handling complex tasks like retrieving banking details or booking travel.

These agents represent the industry's next evolution beyond current digital assistants, Amon said, with form factors ranging widely as companies experiment with always-on devices that see and interpret the surrounding world. The comments arrive as major players including Apple and Samsung face pressure to adapt, while Qualcomm positions itself at the center of a wave of consumer electronics that blend hardware and advanced AI capabilities.

Yet the same technology driving these innovations is colliding with political turbulence in Washington. The Trump administration's recent decision to place Anthropic's advanced Mythos 5 and Fable 5 models under export controls has drawn sharp criticism from foreign governments concerned about reliability of U.S. AI systems. Canadian Prime Minister Mark Carney warned that overreliance on certain American models carries clear risks, noting that sudden policy shifts could leave allies without dependable options. European officials have responded by accelerating their own tech sovereignty efforts, including expanded investment in domestic data centers and semiconductors to reduce dependence on U.S. providers.

The administration's moves follow a pattern of abrupt reversals on AI oversight. An earlier proposal for voluntary reporting on advanced models was delayed, then replaced by a narrower order that avoided mandatory licensing. Critics argue the export restrictions function as de facto licensing requirements, sending mixed signals to global markets at a moment when U.S. dominance in AI is being actively contested.

Compounding these regulatory uncertainties is the rapid spread of AI-generated content in American elections. Attack ads featuring deepfakes have become commonplace, often placing candidates in fabricated scenarios without disclosure requirements. In Texas, a Trump-aligned group produced an ad depicting Democratic Senate nominee James Talarico in a dress performing a song about transgender children, while the National Republican Senatorial Committee earlier used AI to show Talarico reciting fabricated versions of his own social media posts. Similar tactics appeared in Kentucky's GOP primary, including deepfakes of Representative Thomas Massie in compromising situations with Democratic lawmakers, and in Georgia where one gubernatorial candidate released an entirely AI-generated spot.

Democrats have signaled intent to pursue federal disclosure rules if they regain congressional majorities, arguing that the absence of guardrails allows disinformation to proliferate unchecked. The convergence of hardware breakthroughs, export policy volatility and unregulated synthetic media underscores how quickly AI is moving from laboratory promise to contested political terrain, with consequences for both technological leadership and democratic norms.