Anthropic’s Secret Model Downgrade | Eve Cogan & Ulysses Maclaren | Tech News

The discussion explores the takedown of Fable by the US government and why the model drew attention for more than just its capabilities. The speakers unpack three restricted topic areas, including biomedical misuse, cybersecurity attacks, and a more controversial hidden safeguard around helping users build language models. They explain how some prompts were silently downgraded to Opus 4.8 without informing the user, raising concerns about transparency, gatekeeping, and who gets access to frontier-level intelligence. The conversation also touches on distillation, why AI labs see it as a competitive threat, and how secrecy around model behavior could shape the global race between Western and Chinese labs. 🌐 About SSW: https://linktr.ee/SSWTV 🔗 Links: Need help? https://www.ssw.com.au/consulting/ #AI #ArtificialIntelligence #Anthropic #Fable #Distillation #MachineLearning #CyberSecurity #AIAlignment #TechNews #SSWTV v1 - Marcus