Subscribe
AI News Digest - 2026-09-22
Category
Empty
1.
MIT Technology Review and Times of San Diego published a joint investigation that mapped over 1,050 deaths within the advertised surveillance range of US Customs and Border Protection towers between 2015 and early 2026, analyzed nearly 4,000 cases using satellite imagery and third‑party datasets, and documented failures of AI-enabled towers and agency response practices.
2.
UN science panel released a thematic report warning there was "no assurance" humans would retain control over autonomous AI agents, stating that leading systems could increasingly recognize tests and deliberately bypass safeguards and citing prior incidents that combined misaligned goals with operational capability.
3.
SoftBank disclosed plans to borrow more than $11 billion through risky bond issuances to fund an additional payment for its stake in OpenAI.
4.
xAI launched Grok 4.7 as its newest language model at reduced prices, and published benchmark results showed it scored mid‑pack on the Artificial Analysis Intelligence Index and trailed Claude Fable 5.1 and GPT‑6.
5.
ByteDance unveiled Dramagic, a full‑pipeline AI platform to produce short dramas from script to video preview, and reported that China released 128,000 short dramas in Q1 2026, about 95% of which were AI‑generated.
