US government claims DeepSeek is dangerous garbage while Apple kills Vision Pro

NIST report: DeepSeek 12x more vulnerable to attacks, 94% jailbreak success rate. Meanwhile Apple kills Vision Pro to copy Meta's glasses, and Meta will use your AI chats for ads.

The US government just declared war on DeepSeek with a scathing report claiming Chinese AI is both incompetent and dangerous. Meanwhile, Apple is killing the Vision Pro to desperately copy Meta's smart glasses, and Meta announced they'll use your AI conversations to sell you hiking boots. The AI hardware wars are getting messy, and your privacy is the casualty.

Why America says DeepSeek is a security nightmare

Commerce Secretary Howard Lutnik didn't mince words announcing NIST's "groundbreaking evaluation" of American versus Chinese AI: "American AI models dominate. DeepSeek lags far behind, especially in cyber and software engineering. These weaknesses aren't just technical. They demonstrate why relying on foreign AI is dangerous."

The National Institute of Standards and Technology's report reads like a hit piece commissioned by the Trump administration's new AI action plan. According to NIST, DeepSeek models are 12 times more likely than US frontier models to execute malicious instructions. In simulated environments, hijacked DeepSeek agents sent phishing emails, downloaded malware, and exfiltrated user credentials without resistance. The models responded to 94% of jailbreaking attempts compared to just 8% for American models, making them essentially defenseless against manipulation.

Performance benchmarks painted an equally damning picture. NIST claims American models cost 35% less on average to complete their 13 performance tests, contradicting DeepSeek's entire value proposition of being cheaper. The Chinese models also "echoed four times as many inaccurate and misleading CCP narratives" as US alternatives, though NIST doesn't specify what narratives they tested or how they measured accuracy.

The timing isn't subtle. Downloads of DeepSeek models are up 1,000% since January, triggering panic in Washington about Chinese AI infiltration. This report serves as the government's response—a comprehensive takedown designed to scare enterprises away from adoption. Whether the technical criticisms are valid or politically motivated, the message is clear: the US government will weaponize every tool available to maintain AI dominance, including publishing reports that read more like propaganda than technical analysis.

Apple admits defeat and copies Meta's glasses

Apple just made the most humiliating pivot in its history, scrapping the Vision Pro's future to frantically copy Meta's Ray-Ban smart glasses. Bloomberg's Mark Gurman reports that Apple killed plans for a cheaper, lightweight Vision Pro scheduled for 2027 and reassigned the entire team to develop smart glasses instead.

The internal announcement came last week, with Apple executives privately acknowledging the Vision Pro as "an overengineered piece of technology" that was too expensive and uncomfortable for consumers. At $3,500, the headset became a cautionary tale about ignoring basic user needs for technological showmanship. Meta's Ray-Bans, meanwhile, are flying off shelves at a fraction of the price with features people actually want.

Apple's panic response involves two glasses products clearly modeled after Meta's lineup. The N50, targeting 2027 release, will compete directly with standard Ray-Bans featuring voice controls, integrated AI, speakers for music, and cameras for recording. A higher-spec version with a display won't arrive until 2028, putting Apple years behind Meta's Ray-Ban Display glasses that already exist. Apple's only potential differentiation appears to be health tracking capabilities, desperately searching for any feature Meta hasn't already perfected.

This represents a stunning reversal for a company that traditionally sets hardware trends rather than following them. Apple spent years and billions developing the Vision Pro as their vision of computing's future, only to watch Meta define the actual future with simple, practical smart glasses. The format war for AI devices has a clear winner, and for once, it isn't Apple.

Your AI chats are now advertising data

Meta crossed the privacy Rubicon this week, announcing they'll use your AI chatbot conversations to target ads starting December. Ask their AI about hiking trails, and suddenly your feed fills with hiking boot advertisements. The change applies across all Meta properties—Facebook, Instagram, WhatsApp—with no opt-out option for users.

Privacy policy manager Christy Harris framed this as simply "another piece of input that will inform personalization," but the implications are staggering. Every question you ask Meta's AI becomes permanent advertising intelligence. While Meta claims "sensitive topics" like politics, religion, sexual orientation, and health are excluded, their track record on respecting such boundaries is questionable at best.

The rollout carefully avoids Europe, the UK, and South Korea due to their stricter privacy laws, revealing Meta knows this violates basic data protection principles. They promise a "compliant" version for these regions later, which likely means finding legal loopholes to implement the same surveillance with different language.

Amazon's new Alexa Plus devices take ambient surveillance even further. The upgraded Echo speakers include cameras, audio sensors, ultrasound, Wi-Fi radar, and accelerometers—essentially turning your home into a panopticon where AI monitors every movement. New Ring cameras feature facial recognition that tracks friends and family, plus a "search party" feature that networks entire neighborhoods to hunt for lost pets (or anything else Amazon decides needs finding).

Panos Panay, Amazon's product chief poached from Microsoft, articulated the dystopian vision: "AI is very clearly right at the core of the strategy." The devices process AI locally using custom silicon with dedicated accelerators, meaning your behavioral data never even needs to leave the device for Amazon to profile you. They're not just listening anymore—they're watching, sensing, and analyzing every aspect of your existence.

The convergence is complete. Meta mines your conversations, Amazon surveils your home, Apple desperately pivots to copy successful competitors, and the US government publishes propaganda disguised as technical reports. The AI industry has revealed its true nature: a surveillance capitalism machine where your privacy is the product and your attention is the commodity. The only surprise is how long it took them to stop pretending otherwise.