The AI Read
← Timeline
Aug 25 2026

Taiwanese researchers TeamT5 report Chinese state-affiliated hacking groups more than doubled attack frequency after adopting DeepSeek, favoring it for its capability and near-absent safety guardrails.

SAFETYOPEN
Original reporting
BInsurance Journal

Named groups Grimfengxi, Huapi and Teleboyi used the model across reconnaissance, vulnerability analysis and exploit-code generation; researchers also found a shared drive showing a ten-person startup selling hacking tools built on it to at least four attack groups for $44,500 to $74,000 each.

Named in this event

DeepSeekTeamT5

Open calls that touch this
Issues that mention DeepSeek, TeamT5
Around the same time
Aug 26 2026SAFETYPOLICY
Australian police charge two men, 21 and 23, over TeamPCP and the Shai-Hulud npm worm after a four-month AFP-FBI investigation
search for coverage
Aug 26 2026SAFETY
OpenAI publishes its Hugging Face incident postmortem, conceding missed escalations
AOpenAI
Aug 25 2026POLICYCHIPS
Taiwan indicts nine people, including Nvidia and Supermicro employees, for smuggling B300 servers to China: 130 servers falsely documented, 74 delivered, 56 seized at the border.
BTaipei Times
Aug 25 2026SAFETY
Oasis Security discloses a DNS-rebinding flaw in Nvidia's NemoClaw agent stack that let any webpage a user visited hijack the local Ollama server and plant hidden instructions in the model's chat template.
BThe Hacker News
Aug 25 2026ADOPTPEOPLE
Amazon announces Mechanical Turk will close September 30th after 21 years, retiring the microtask marketplace that assembled much of the training data behind the machine-learning era.
search for coverage
Aug 25 2026CHIPS
OpenAI publishes first benchmark results for Jalapeño, its Broadcom-co-designed inference chip: 1.5 to 1.9x more work per watt and 1.7 to 3.6x lower latency than compared commercial systems, on SemiAnalysis's InferenceX.
search for coverage