United States uncovers China-based anti-AI data campaign | JQJO

The New York Knicks celebrated a championship that ended a fifty‑three‑year title drought. During the opening‑night ceremony, players received rings, but assistant coaches were handed empty boxes. Many Madison Square Garden employees did not receive rings at all. In September, the organization sent a phishing‑style email asking employees to choose a ring style; those who clicked were required to attend a cybersecurity class. The incident highlights the Knicks’ ongoing focus on security and the uneven distribution of championship honors within the franchise.

Mikita Balesni and Tomek Korbak, two former OpenAI safety researchers, gave their first interview since their dramatic dismissal. They discussed the company’s internal investigation into the Hugging Face hack and questioned the logic of OpenAI’s rebuttal. The interview also highlighted ongoing tensions over AI safety within the organization.

China’s Communist Party will convene a closed‑door plenum from Monday to Thursday, gathering about three hundred senior members to discuss tightening discipline. The session, the fifth since the current Central Committee was formed in 2022, is seen as a chance for President Xi Jinping to reinforce his authority and sustain the anti‑corruption campaign he launched in 2014. Analysts expect the meeting to signal system‑wide changes rather than personnel moves, as Xi’s purges have already investigated millions of officials and cleared hundreds of senior leaders. The plenum is likely the last major gathering before the 2027 leadership reshuffle, when many of Xi’s close associates are expected to retire.

Google Chrome now offers an Enhanced Protection mode that warns users when they visit sites resembling known scams. The new "suspicious site" alerts appear as smaller pop‑ups, offering options to mark a site as safe or return to safety. Unlike full‑page dangerous site warnings, these alerts indicate potential risk based on patterns similar to malicious sites. Users enable the feature in Chrome settings under Safe Browsing, which also uses AI analysis in real time. The feature requires sharing more browsing data with Google, linked to the user’s Google Account.

A security researcher demonstrated a software‑only jailbreak of the Switch 2, gaining userland access that allows running homebrew, backing up save files, and accessing most peripherals. The exploit relies on an integer overflow in the original Switch version of SmileBasic 4, a programming environment previously used in the 3DS smilehax hack. After about two weeks of testing, the researcher discovered a 32‑bit overflow that exposes an array pointer, enabling read/write access to system memory. From there, a buffer overflow in the FileSystem services module granted arbitrary read/write privileges, leading to native code execution. The proof‑of‑concept included modded characters in Mario Kart World, but the current userland exploit remains sandboxed and does not provide kernel‑level control.