Tech News — 2026-04-05#

Story of the Day#

Suspected North Korean hackers deployed an elaborate AI deepfake scheme masquerading as tech founders to trick top open-source maintainers. The attackers successfully compromised widely used Node.js tools like Axios, injecting self-destructing malware into the supply chain before developers even noticed.

Top Stories#

[Top NPM Maintainers Targeted with AI Deepfakes in Massive Supply-Chain Attack] · Slashdot North Korean hackers deployed AI video deepfakes to trick top open-source developers into executing remote access trojans via fake virtual meetings. The coordinated attack briefly compromised the axios package—which is downloaded millions of times daily—sneaking in malware that erased itself and left a clean decoy within seconds. It is a terrifying escalation in social engineering that proves our web infrastructure is only as secure as the humans maintaining it.

[Samsung will discontinue its Messages app in July and replace it with Google’s] · Engadget Samsung is finally putting a bullet in its proprietary SMS app this July, pushing all US Galaxy owners to switch to Google Messages. The move gives users default access to cross-device RCS features like read receipts and high-res media, plus Google’s Gemini-powered generative AI tools for remixing photos. This marks the quiet, long-overdue end to a heavily fragmented messaging era for Android.

[The UK government reportedly wants Anthropic to expand its presence in London] · Engadget The UK government is aggressively courting AI darling Anthropic with proposals for a London office expansion and a potential dual stock listing. This opportunistic push follows Anthropic’s public falling out with the US Department of Defense, which temporarily designated the startup a “supply chain risk” after the company refused to budge on specific AI safety guardrails. It’s a high-stakes geopolitical scramble over AI sovereignty, though Anthropic will still have to compete with OpenAI’s established London footprint.

[Claude Code Leak Reveals a ‘Stealth’ Mode and a ‘Frustration Words’ Regex] · Slashdot A massive leak of Anthropic’s Claude Code source code has exposed some bizarre internal mechanics, including an “undercover mode” that allows the AI to make stealth contributions to public code bases. Hilariously, the code also contains a regex tool explicitly designed to scan user prompts for expletives and other colorful signs of human frustration. Why exactly an enterprise LLM needs to precisely calculate how angry you are remains an open question.

[Scientists Engineered a Plant To Produce 5 Different Psychedelics At Once] · Slashdot Researchers have successfully modified a tobacco plant using genes from mushrooms, toads, and other plants to simultaneously synthesize five different natural psychedelics, including DMT, psilocybin, and 5-MeO-DMT. As therapeutic interest in these compounds surges, this engineered crop could bypass the ecological and ethical nightmares tied to harvesting wild organisms like the Sonoran Desert toad.

[Will ‘AI-Assisted’ Journalists Bring Errors and Retractions?] · Slashdot A single “AI-assisted” reporter at Fortune generated over 600 articles in six months using tools like Perplexity, personally accounting for roughly 20% of the publication’s web traffic. The highly efficient pipeline allows writers to churn out rewritten press releases in minutes, prompting media companies like USA Today to openly hire for AI-assisted reporting roles. Despite union leaders warning against the loss of human judgment and expertise, executives are treating the pivot as an inevitable cost-saving measure.

Also Worth Knowing#

• [Grammarly’s sloppelganger saga] (The Verge): The omnipresent spell-check browser extension is pivoting to a broader mission, rebranding as an AI platform called Superhuman.
• [Does Ubuntu Now Require More RAM Than Windows 11?] (Slashdot): Canonical’s latest Ubuntu 26.04 LTS release bumps its minimum baseline to 6GB of RAM, quietly admitting that the bloated reality of modern web browsing demands heavier hardware.
• [Suno is a music copyright nightmare] (The Verge): The AI music generator’s copyright filters are proving laughably easy to bypass, allowing users to quickly spit out near-perfect clones of tracks by artists like Beyoncé and Black Sabbath.
• [Copilot is ‘for entertainment purposes only’] (TechCrunch): In a stark admission of its own product’s unreliability, Microsoft’s terms of service now warn users not to trust its AI, labeling Copilot “for entertainment purposes only”.
• [CBP facility codes sure seem to have leaked via online flashcards] (Ars Technica): Highly confidential security procedures and facility codes for US Customs and Border Protection in Texas were accidentally left exposed to the public in a Quizlet flashcard study set.