Click any tag below to further narrow down your results
Links
Five Eyes intelligence agencies warn that frontier AI models able to mount complex cyber attacks will emerge in months, lowering barriers for bad actors. They urge treating cyber risk as a core business and societal responsibility, citing the US block on foreign use of Anthropic’s Fable and warning of other advanced models in development.
- Five Eyes intelligence agencies warn AI models capable of devastating cyber attacks on governments and businesses will emerge within months, drastically lowering the barrier for bad actors.
- The US has already barred foreign nationals from using Anthropic's Fable and Mythos models, citing national security concerns over their ability to find and exploit security flaws.
- Australia has signed a non-binding deal with Anthropic to share AI progress, favoring a "light-touch" regulatory approach to capture economic benefits despite the risks.
- Experts warn other states or companies, including China, could soon develop similar or more advanced offensive AI systems.
@Rafa_Schwinger on X
The post argues that Anthropic's Claude Mythos and Fable models owe their edge to verifiable reinforcement learning signal rather than a secret architecture. The author, writing under personal inference, says the lead is measured in months because competitors can replicate the recipe.
- Capability is modeled as a strong base foundation multiplied by gradeable signal, meaning tasks whose outcomes can be checked cheaply, like code that compiles or passes tests.
- Long-horizon endurance depends on keeping per-step errors low and managing context tightly; METR puts Mythos's 50% success horizon at 16+ hours, though measurements past that are unreliable.
- The same verifiability that makes the model strong at coding and security also makes it effective at finding vulnerabilities, with partners reporting roughly 1,500 independently confirmed findings in the first month.