TECH
US government blocks Anthropic’s top AI models amid security concerns
The Verge reports the White House ordered Anthropic to cut off foreign access to its newly released Fable 5 and Mythos 5 models over fears they could be used for cyber‑attacks and may have been accessed by China, sparking a sovereignty debate. TechCrunch contends the ban stemmed from an undisclosed export‑control directive and political friction rather than a proven jailbreak, warning it sets a precedent for government control of U.S. AI. Both outlets note the shutdown forced Anthropic to suspend the models for all users.
Meta launches AI mode search and creative tools on Facebook, Instagram
According to The Verge, Meta’s new AI mode appears alongside standard search options on Facebook and generates answers from publicly posted content across its apps. Powered by the Muse Spark model, the feature lets users ask follow‑up questions and is paired with AI‑generated photo presets and collage suggestions for Instagram. Meta says the tools will expand over time to cite recommendations and content from Instagram, Facebook and Threads.
Xbox shuts down Ninja Theory and plans studio spin‑offs amid restructuring
The Verge reports that Microsoft is closing Ninja Theory, the studio behind the Hellblade series, after staff were notified on a call. Bloomberg and Kotaku say other Xbox studios, including Compulsion Games and Double Fine, are in active negotiations to spin off. The actions follow Xbox CEO Asha Sharma’s recent remarks about a “reset” and come amid broader executive departures at the division.
Big tech seeks AI preemption as Congress debates child‑safety bill
The Verge reports that major technology firms are lobbying for a single federal AI preemption law, attempting to attach it to the Kids Online Safety Act (KOSA). Disagreements between Senate and House versions of KOSA and the exclusion of House Republicans have created confusion and skepticism among lawmakers, child‑safety advocates, and industry insiders. With a packed legislative calendar and upcoming midterms, critics doubt the combined bill will advance before Congress recesses.
Chinese Zhuque‑2E rocket breakup creates debris near Starlink constellation
Ars Technica reported that the upper stage of China’s Zhuque‑2E launch shattered in low‑Earth orbit on June 9, producing an estimated 100‑150 fragments that now orbit between 335 and 424 km, crossing the path of the ISS and hundreds of Starlink satellites. The US Space Force confirmed the breakup and incorporated the objects into routine conjunction assessments, saying there is no current danger to crewed missions. Aerodynamic drag should cause most debris to re‑enter the atmosphere within months, though analysts note the incident adds to a growing Chinese contribution to long‑lived orbital junk.
Google to remove legacy ad‑blocker support in upcoming Chrome updates
According to The Verge, Chrome versions 150 and 151 expected later this summer will eliminate the final workarounds that let older ad blockers such as uBlock Origin operate. The changes remove references to the deprecated Manifest V2 extension platform, meaning only Manifest V3‑compatible blockers will function in Chrome 151 onward. Google says the move reduces technical debt and security risks, after having phased out MV2 support in 2024.
Nvidia aims to raise $25 billion in first bond sale since 2021
The Financial Times reported that Nvidia is launching a seven‑part, investment‑grade bond offering that was upsized to $25 billion after receiving $85 billion in orders, with a 10‑year tranche yielding about 0.5 percentage points over U.S. Treasuries. The proceeds will be used for general corporate purposes, including repayment and refinancing of existing notes. Ars Technica emphasized that the deal tests investor appetite for further exposure to the AI sector and noted concerns about concentrated risks within the AI financing ecosystem.
AMD removes memory encryption from consumer CPUs, sparking user backlash
AMD silently disabled Transparent Secure Memory Encryption (TSME) on its consumer Ryzen processors through a recent AGESA firmware update, limiting the feature to PRO and EPYC chips. Users, including a Linux hobbyist who detected the change with Host Security ID, reported the loss of the feature and criticized AMD’s vague statement that TSME is only for PRO CPUs. The lack of notice and unclear explanation has prompted backlash over the perceived reduction in security for standard consumer hardware.
Salesforce to acquire AI customer service platform Fin for $3.6 billion
Salesforce announced it will buy Fin, the AI‑driven customer service platform formerly known as Intercom, for $3.6 billion. The acquisition aims to bolster Salesforce’s Agentforce suite with Fin’s multi‑channel AI agents that handle chat, WhatsApp, SMS, phone calls and Slack. The deal is slated to close in the last quarter of Salesforce’s 2027 fiscal year.
Chinese hackers infiltrated US medical and military networks for over a year
The Register reported that a Chinese state‑linked espionage group, tracked by Google as UNC6508, maintained persistent access to multiple North American medical and military research institutions for more than a year. Using custom malware called InfiniteRed, the actors harvested REDCap credentials and set up a Gmail compliance rule to exfiltrate emails about defense, drone technology and the Chikungunya virus. Google’s threat‑intelligence team identified the campaign, alerted the affected organizations and disabled the attacker‑controlled Gmail account.