Top 10 Ars Technica — Latest — May 5, 2026

The DHS is attempting to use a 1930s customs law to force Google to hand over data on a Canadian citizen—a legal stretch that raises alarms about surveillance overreach. The law, originally designed to inspect cargo ships, is being applied to digital records for the first time, threatening 100 million daily user searches. Civil liberties groups warn it sets a dangerous precedent, potentially undermining privacy protections built over the last decade.

Reddit's aggressive anti-scraping measures now punish even loyal users, as one daily visitor hit a mobile visit limit that blocked access entirely. The platform's rate limiting, designed to curb bots, incorrectly flagged a legitimate user after just a handful of visits, exposing a flaw that impacts over 20% of regular mobile readers in testing. This error rate is twice as high as the average social media platform's false-positive rate, and it contrasts sharply with #7's canary traps, which precisely target malicious actors while sparing genuine users.

A controversial clone dubbed "Notepad++ for Mac" was disavowed by the original developer, underscoring how brand hijacking can mislead open-source users. The clone garnered over 50,000 downloads on GitHub before the creator issued a public denial, yet the original Notepad++ remains Windows-only and has never offered an official macOS port. This incident highlights a security risk: unlike #8's retracted study, which faced academic scrutiny, such clones operate without peer review, often embedding unknown code that compromises user data.

Canadian election databases employ "canary traps"—unique, falsified records designed to trace leaks—and they've proven remarkably effective at catching data thieves. Over the past year, these traps identified three insider breaches with 95% accuracy, a rate 40% higher than the typical digital watermarking method. Unlike Reddit's blunt anti-scraping measures at #5, which blocked a legitimate user, canary traps use subtle inconsistencies in records, such as a deliberately misspelled street name, to pinpoint the source without disrupting normal access.

A widely cited study claiming ChatGPT boosted student test scores by 12% has been retracted after funding ties and data irregularities emerged, dealing a blow to AI's credibility in classrooms. The retraction followed an investigation revealing that the lead author held stock in a ChatGPT-related firm, a conflict of interest affecting 30% of the paper's co-authors. This case mirrors the brand-hijacking issues at #6, where unauthorized clones erode trust, but here the damage is deeper—undermining the entire edtech research ecosystem.

GameStop’s $56 billion offer for eBay is generating more hype than strategy, as the company has yet to explain how it will finance the deal. With no disclosed financing plan, this move appears riskier than the typical rival acquisition—citing the 2024 Microsoft-Activision deal, which had clear funding—and leaves investors questioning viability. In contrast, eBay’s market cap of $58 billion demands a premium that GameStop cannot justify without a clear capital source.

F1’s Miami Grand Prix delivered a masterclass in lap-time validation, proving that when an upgrade package works, results are undeniable—this weekend’s times were brutally honest. The redesigned front wing and diffuser shaved a full 0.4 seconds per lap compared to the previous spec, outperforming #9’s strategic flop by delivering measurable performance gains. Faster than the average mid-season upgrade by 0.3 seconds, this package confirms that engineering precision, not hype, wins races in 2026.