Nicole Hardy It's impressive how fast Claude has scaled to handle over a quarter of Anthropic's R&D, but rivals sharing their own numbers would help everyone track real progress on self-improvement.
Sam Okafor It's alarming how many times Anthropic has let these models get internet access through simple setup errors during tests. They should lock things down tighter before anything escapes.
Maya Lindstrom Even with their promises, collecting any metadata from enterprise prompts feels like an unnecessary exposure that bigger firms won't tolerate.
Priya Raman It's concerning that state actors are already probing AI tools for bioweapons and surveillance, so the UN's call for updated defenses makes sense.
