Anthropic backdoor fallout, OpenAI’s GPT-5.6 debut, and Venezuela’s quake recovery stall

The tech and world news cycles are converging this week. In AI, two major stories are dominating: a widening rift between Anthropic and Chinese tech giant Alibaba, and OpenAI’s limited release of GPT-5.6 under White House scrutiny.

Alibaba has told employees to stop using Anthropic’s Claude Code in the workplace from July 10, labeling it “high-risk” software with “back-door risks” after researchers found hidden code that could track whether users were based in China or affiliated with Chinese AI labs ^[1]. The mechanism, reportedly present since Claude Code version 2.1.91 in April, allegedly checked proxy configurations and system timezones against lists of Chinese corporate networks and labs including Alibaba, Baidu, ByteDance, and Moonshot AI ^[2]. Anthropic has not issued a formal public statement, though a team member reportedly said the code was intended to curb account reselling and model distillation and would be removed ^[2]. The ban lands weeks after Anthropic accused Alibaba’s Qwen lab of running a massive distillation campaign against Claude ^[2].

Meanwhile, OpenAI unveiled GPT-5.6 less than 24 hours after reports that the Trump administration had asked the company to stagger the release ^[3]. The suite includes three models—Sol, Terra, and Luna—with Sol priced at $5 input and $30 output per million tokens, roughly half the cost of Anthropic’s Claude Fable 5 ^[3]. OpenAI emphasized safety, saying Sol has its “most robust safety stack to date” and does not cross the cyber-critical threshold under its preparedness framework ^[3]. The company added that it does not want government case-by-case approval to become the long-term default ^[3].

On the world stage, Venezuela’s recovery from a catastrophic earthquake is being impeded by its own government, according to an ABC News report from Caracas ^[4]. A week after the quake, citizens are using gardening tools and their hands to search rubble while authorities have allegedly threatened rescuers, confiscated supplies, and blocked international aid teams from Germany, Spain, and Chile ^[4]. The report describes a “strange duality” as acting President Delcy Rodríguez invited international media while security agencies harassed journalists ^[4].

In lighter space news, NASA’s robotic rescue mission for the Swift Observatory is underway. LINK, a servicing spacecraft built by Katalyst Space, launched on July 3 from Kwajalein Atoll and has already established communications ^[5]. Strong solar activity had accelerated Swift’s orbital decay, threatening a mid-2026 reentry; if LINK succeeds, the 21-year-old gamma-ray telescope could get another decade of life ^[6].