This week’s newsletter dives into the intensifying AI arms race, unprecedented regulatory showdowns, and the growing collision between innovation and national security. With three high-stakes hearings on trade secrets, defense AI posture, and America's competitive edge in computing, plus major developments in data privacy and surveillance tech, you won’t want to miss what’s shaping the next chapter in tech policy.

In this week’s Nimitz Tech:

• DOGE: Elon Musk’s startup-style government is using AI not to streamline services—but to systematically gut them.

• Halicinations: AI bots are getting smarter at math—but dumber at facts—and even their creators can’t explain why.

• Accountability: TikTok just got hit with a $600 million EU fine for quietly sending user data to China—and regulators say more penalties may follow.

WHO’S HAVING EVENTS THIS WEEK?

Wednesday, May 5th

• 🔐 House Hearing: “Protecting Our Edge: Trade Secrets and the Global AI Arms Race.” House Committee on the Judiciary, Subcommittee on Courts, Intellectual Property, Artificial Intelligence, and the Internet. Hearing scheduled for 10:00 AM in 2141 Rayburn HOB. Watch here.

Thursday, May 6th

• 🪖 House Hearing: “Information Technology and Artificial Intelligence Posture of the Department of Defense.” House Committee on Armed Services, Subcommittee on Cyber, Information Technologies, and Innovation. Hearing scheduled for 9:00 AM in 2212 Rayburn HOB. Watch here.

• 💻️ Senate Hearing: “Winning the AI Race: Strengthening U.S. Capabilities in Computing and Innovation.” Senate Committee on Commerce, Science, and Transportation. Hearing scheduled for 10:00 AM in 253 Russell SOB. Watch here.

TECH NEWS DRIVING THE WEEK

In Washington

• The Department of Government Efficiency (DOGE), Elon Musk’s chaotic attempt to run the U.S. government like a startup, is aggressively integrating artificial intelligence across federal agencies—but not with caution or clarity. While some applications, like using AI to flag regulatory overreach at HUD, have limited practical merit, others aim to replace tens of thousands of public workers with immature AI agents, despite their known unreliability. Rather than improving public services, DOGE’s approach reflects a reckless effort to dismantle the administrative state under the guise of innovation, often obscuring data usage and oversight. The problem isn’t AI itself, but DOGE’s unaccountable deployment of it to hollow out public institutions in ways that benefit private tech interests.

• President Trump’s fiscal 2026 budget proposal aims to dismantle the Cybersecurity and Infrastructure Security Agency’s (CISA) disinformation offices, accusing them of colluding with Big Tech to censor conservative voices and misusing agency resources. Labeling the offices as part of the so-called “Censorship Industrial Complex,” the administration seeks to slash CISA’s budget by nearly $500 million—about 16%—and eliminate what it calls “weaponized” spending. The plan reflects a broader effort to shrink the federal government and reframe CISA’s role around defending federal networks and physical infrastructure, rather than combating misinformation. The move follows years of Republican attacks on CISA’s work related to election integrity and the firing of former director Chris Krebs, whose security clearance has now been revoked amid new investigations.

• Amid growing concerns over AI chip smuggling to China, U.S. Rep. Bill Foster is preparing new bipartisan legislation that would require companies like Nvidia to embed location-tracking and export control enforcement into their AI chips. Prompted by reports of banned chips fueling China’s AI development—including military applications—the bill would mandate the U.S. Commerce Department to issue rules within six months. The proposed measures include verifying chip locations post-sale and disabling chips not authorized under export controls. Experts note the tracking technology already exists and is partially built into current Nvidia chips, with similar systems reportedly in use by Google. With national security at stake, both Democratic and Republican lawmakers support the move as a needed step in closing loopholes in U.S. export enforcement.

National

• Despite advances in artificial intelligence, hallucinations—when AI systems fabricate information—are becoming more frequent, especially in new “reasoning” models from companies like OpenAI, Google, and DeepSeek. These models are designed to solve complex tasks step by step, but the increased complexity appears to amplify their tendency to invent facts, with some tests showing hallucination rates as high as 79%. Companies are struggling to explain or fix the issue, as the systems are trained on vast, opaque datasets and rely on probabilistic outputs rather than grounded truth. Techniques like reinforcement learning have improved performance in tasks like math and coding but seem to degrade factual reliability, particularly when bots attempt to mimic human reasoning. The rise in hallucinations is causing growing concern, especially for applications involving sensitive or factual information, as users are left to sort truth from fiction in AI-generated responses.

• Meta has built a massive data classification and privacy infrastructure known as Privacy Aware Infrastructure (PAI) to address the growing complexity and scale of its data ecosystems. Through a “shift-left” approach, the company embeds data schematization and annotation directly into the development process, supported by a universal privacy taxonomy that enables consistent labeling across systems. Tools like OneCatalog track and manage millions of data assets in real time, while machine learning models and heuristics power high-precision classification at scale. The effort aims to ensure regulatory compliance and user privacy while improving developer efficiency and enabling new product features, as illustrated by the granular data protections built into features like religious views in Facebook Dating. Ultimately, Meta’s strategy represents a highly technical, tightly integrated, and collaborative approach to managing data privacy at global scale—transforming compliance from a constraint into a catalyst for innovation.

• Artificial intelligence is rapidly advancing in its ability to identify precise locations from seemingly innocuous photos or sounds, as demonstrated by recent tests involving ChatGPT and Perplexity. These tools accurately geo-guessed photos even when stripped of metadata, using subtle visual cues like architecture, landscape, or even the brand of a wheelbarrow to pinpoint locations like Cork, Ireland or rural Netherlands. The same principle extends to sound: when fed a list of bird species identified through an app, the AI models were able to estimate the user's location to a specific region in Northern Europe. This growing precision raises serious privacy concerns, as public content—photos, videos, and audio—can unintentionally disclose location data. With social media content fueling training datasets, AI’s location inference capabilities are poised to become even more accurate, making situational awareness more crucial than ever for users who value anonymity.

International

• The European Union has fined TikTok €530 million ($600 million) for illegally transferring user data to China, following a four-year investigation by Ireland’s Data Protection Commission. The regulator found TikTok violated the EU's strict GDPR privacy rules by failing to disclose or safeguard data accessed remotely by staff in China, raising serious concerns about surveillance risks. TikTok argued the decision ignored recent security investments, including its Project Clover data localization initiative, and said it would appeal. The company also faces further scrutiny for providing inaccurate information during the probe, including a delayed admission that some European user data had been stored on Chinese servers. The case underscores the EU’s hardening stance on data sovereignty and the growing pressure on tech companies to keep sensitive user data within regional legal protections.

FOR FUN

• 💃 Swing Dance Celebration at the Embassy of France (5/8)

ADD TO THE NIMITZ NETWORK

Know someone else who would enjoy our updates? Feel free to forward them this email and have them subscribe here.

Additionally, if you are interested in our original publication on veterans affairs policy, check it out below: