SecurityHacktron: HEIF and SSO Chain Reached OpenAI Code RepositoriesResearchers chained a HEIF decoder flaw with an SSO weakness. Claude accelerated exploit development but did not act autonomously.September 21, 2026·2 min read▲—▼
AI & ChipsAnthropic Weighs New Model as Safety Pacing Meets IPO PressureAnthropic is considering a model release to counter Astra's momentum, Reuters reports. Neither release nor timing is confirmed.September 19, 2026·2 min read▲—▼
AI & ChipsAnthropic Confirms Its Own Wet Lab for AI-Assisted BiologyAnthropic runs a wet lab where Claude can support physical experiments, while drug discovery and clinical trials remain separate.September 18, 2026·2 min read▲—▼
SecurityAnthropic: AI Agents Automate Attacks — Attribution Still Comes From the VendorAnthropic documents agentic attacks and 151 million exchanges attributed to Alibaba. The attribution is the vendor's own.September 11, 2026·2 min read▲—▼
AI & ChipsGerman articleClaude formalisiert Fermats letzten Satz in LeanAnthropic meldet den ersten vollständigen, maschinengeprüften Lean-Beweis. Claude arbeitete elf Tage; die Mathematik selbst ist nicht neu.September 5, 2026·1 min read▲—▼