Daily Tech & GNSS News Digest - June 30, 2026

Tuesday’s news split neatly between two questions the industry keeps circling back to: who controls the model, and who controls the chips underneath it. Anthropic pushed AI further into the physical sciences with a research workbench built for labs, while Meituan answered Washington’s export controls with an enormous open-source model it says never touched a Western chip. The Supreme Court, meanwhile, drew a privacy line around the location data all of this infrastructure quietly collects. On the positioning side, the Army funded a card built to keep working when GPS is jammed, and Europe’s Galileo constellation finished quietly wiring a signal built for everything from smartwatches to smart streetlights.

Tech News

Anthropic Launches Claude Science, an AI Workbench for the Lab

Anthropic released Claude Science in beta on June 30, an AI workbench aimed at scientists that pulls together more than 60 scientific databases, code tools, and compute researchers otherwise juggle by hand, per Bloomberg and STAT News. The tool can automate multistep biology and chemistry tasks—including predicting protein structures—by letting researchers ask questions in plain language instead of querying dozens of individual sources.

Anthropic is leaning on reproducibility as the selling point: every result ships with the exact code and environment that produced it, a plain-language explanation of the method, and the full message history behind it. The beta is available on macOS and Linux for Pro, Max, Team, and Enterprise plans, with discounted seats for academic and nonprofit labs, and the company is funding up to 50 “AI for Science” projects with as much as $30,000 in compute credits each, applications open through July 15. A neuroscientist at the Allen Institute told Anthropic a multi-agent literature-review template built on the tool cut a process that used to take up to two years down dramatically—an early sign of what the company is betting the product can do at scale.

China’s Meituan Open-Sources a Trillion-Parameter Model Trained Entirely on Domestic Chips

Meituan released LongCat-2.0, a 1.6 trillion-parameter open-source model the company says is the largest ever trained end-to-end—both pretraining and inference—on Chinese-made processors, according to Reuters, SiliconANGLE, and Cybernews. The model runs a one-million-token context window and performs comparably to Google’s Gemini 3.1 Pro, trained on a 50,000-chip domestic cluster coordinated through Huawei’s HCCL interconnect system.

The distinction matters because most Chinese models that claim chip independence still lean on Western silicon for the far more demanding pretraining stage, using domestic chips only for inference. LongCat-2.0’s backers say this one didn’t—a direct answer to US export restrictions on advanced semiconductors, and a marker of how quickly China’s domestic chip stack has matured under pressure. It lands the same week South Korea pledged over half a trillion dollars to its own chip and data-center buildout, underscoring how much of the AI race is now being fought over silicon sovereignty rather than model architecture alone.

Supreme Court Rules Police Need a Warrant for Sweeping Phone Location Data

The US Supreme Court ruled 6-3 on June 29 that broad “geofence” sweeps of cellphone location data constitute a search under the Fourth Amendment, requiring a warrant, per Bloomberg, CNN, and NBC News. Geofence warrants let police pull data on every phone that passed through an area during a given window, without targeting a specific suspect—a tool Google and other platforms have handled thousands of times.

Justice Elena Kagan wrote the majority opinion in the case, which arose from a 2019 Virginia bank robbery conviction that leaned on geofence data from Google. The Court rejected the Trump administration’s argument that no warrant should be required at all, but stopped short of ruling on the legality of the specific search, sending that question back to the lower courts. It’s the clearest signal yet that the justices are willing to extend Fourth Amendment protections to the location-data trail modern phones and apps generate by default.

Additional Headlines:

Update: chip-equipment stocks jump on Samsung and SK Hynix investment plans: A day after South Korea’s $576 billion chip pledge, ASML closed up 6.8% in Amsterdam to a fresh record, with Applied Materials and KLA both gaining more than 5% in New York, per Bloomberg—the clearest market read yet on how much equipment demand the memory giants’ new fabs are expected to generate.
Call-center stocks slide on AI customer-service fears: Concentrix and Teleperformance shares fell sharply June 30 as investors bet AI is increasingly capable of handling customer-service work that once required outsourced human agents, per Bloomberg.
Wultra raises €6.8 million for post-quantum authentication: The Czech cybersecurity startup landed a Series A led by Seventure Partners to build post-quantum digital-identity tools for banks, fintechs, and Europe’s forthcoming digital identity wallet, per Tech Startups.

GNSS News

Pacific Defense Wins Army Contract to Build Jam-Resistant PNT Cards

Pacific Defense announced June 29 that it won a contract from the US Army’s Program Manager for Positioning, Navigation and Timing to develop the CMFF APNT Block 2 plug-in card, a rugged navigation module built to keep working when GPS is degraded or denied, per the company’s release and Inside GNSS. The card follows the SOSA and CMOSS open-architecture standards and fuses GPS, alternative navigation signals, inertial sensors, and vehicle data to maintain position and timing when traditional GNSS is jammed.

The award builds on Pacific Defense’s earlier Block 1 card and covers a multi-phase, multi-year development effort across sites in Cedar Rapids, Mukilteo, Sunnyvale, and El Segundo, with initial prototypes destined for Army ground and aviation integration testing. The company says the compact, low-power card is designed for rapid reuse across allied ground, air, and maritime platforms—another sign that “assured PNT” has become a standard line item in next-generation defense procurement, not a bolt-on afterthought.

Galileo’s Low-Power Signal for IoT Devices Completes Its Rollout

The European Space Agency confirmed that Galileo’s new E5a Quasi-Pilot signal component finished deployment across 12 satellites between November 2025 and April 2026, and is now live and free to use in new and upgraded chipsets, according to GPS World and ESA. The signal is built for mass-market, low-power receivers—smartphones, smart-city sensors, and IoT “snapshot” devices that only need a brief GNSS fix—cutting acquisition time by roughly a factor of three and computational load by a factor of eight versus standard Galileo signals.

Because E5a-QP lets receivers acquire a position using only the E5 band instead of leaning on E1, it also improves resilience against spoofing and jamming by removing a single point of dependency. ESA says the rollout is just the opening step: Galileo’s Second Generation satellites will carry expanded Quasi-Pilot signals across additional frequencies, extending the same low-power approach deeper into the IoT and smart-device chipsets that are quietly becoming GNSS’s largest user base by volume.

Key Takeaways

The AI race is splitting into models and silicon: Anthropic’s Claude Science pushes AI deeper into scientific research, while Meituan’s 1.6-trillion-parameter LongCat-2.0—trained entirely on domestic Chinese chips—shows how far chip self-sufficiency has advanced under US export controls.
Location data just got a constitutional guardrail: The Supreme Court’s 6-3 ruling requires a warrant for sweeping geofence data pulls, extending Fourth Amendment protection to the location trails phones generate by default.
Assured PNT keeps professionalizing on both ends of the spectrum: Pacific Defense’s Army-funded card hardens military platforms against GPS jamming, while Galileo’s newly completed E5a Quasi-Pilot signal brings the same resilience thinking to everyday low-power IoT chipsets.