🤖 【AI Daily News】2026.08.16

• Ars Technica: Lawyers suspect courts are using AI, embedding prompts in filings
The judge warns that the complainant parties’ misuse of chatbots can trigger cascading problems【Strategy】

• Anthropic releases multi-agent experiment: finds agent collaboration, collusion, and sabotage
The research team observed emergent failures and novel risks in a Claude Agent cluster【Research】

• IEEE Spectrum: AI can now independently design functional viruses
Gene sequences written by AI bring medical prospects and dual biosecurity concerns【Hot】

• arXiv: A new study reveals that LLM gender bias depends heavily on the way questions are phrased
The same model’s responses to different wording show significantly amplified deviations【Research】

• Netflix launches GenRec: using an LLM to replace the traditional recommendation pipeline
The LLM-native recommendation architecture has already rolled out to multiple regions with low-traffic phased testing【Production】

• CNBC: Exec departures at OpenAI raise a red alert for IPO valuation
C-suite personnel upheaval combined with regulatory pressure makes investors reassess【Investment】

(6 items total)