Anthropic's Claude Opus 4.8 launches as OpenAI faces false flag scandal
Claude Opus 4.8 tops benchmarks while OpenAI PACs allegedly incite violence.
This week's AI news cycle was dominated by Anthropic's launch of Claude Opus 4.8, which brings incremental gains in reasoning and general performance over Opus 4.7. The model tops the Toloka Arena and is now considered a clear daily driver by many in the AI community. Alongside the model, discussions around model welfare (the ethics of AI rights) gained traction. In policy, Trump's revived Executive Order officially imposes prior restraint on frontier model releases, classifying testing processes and shifting responsibility to the NSA rather than CAISI—a move that could centralize power in intelligence agencies.
Most damning, however, is the revelation that OpenAI's political action committee engaged in false flag advocacy, including calls for violence. This deceptive tactic undermines trust in the company, especially as it simultaneously publishes a well-received policy blueprint. Other notable stories include DeepSeek V4 dropping prices permanently (with a catch), Salesforce introducing AI agent goals, and a growing wave of AI-generated songs on platforms like YouTube. Anthropic also filed for its S-1 (going public), while Google raised $84B in compute-related funding. The week underscores a bifurcation: technical progress in AI capabilities versus dangerous political manipulation by leading firms.
- Claude Opus 4.8 offers incremental improvements over Opus 4.7 and tops the Toloka Arena benchmark.
- Trump's Executive Order classifies AI testing and tasks the NSA with oversight, raising civil liberties concerns.
- OpenAI's PAC reportedly engaged in false flag advocacy for violence, damaging the company's credibility.
Why It Matters
AI regulation tightens via prior restraint, while deceptive political tactics threaten public trust in AI companies.