Who Does Anubis Stop? (fzakaria.com)

A recent blog post by Farid Zakaria sheds light on the challenges posed by Anubis, an HTTP proxy implementing proof-of-work requirements to control access to online resources. While Zakaria utilized a large language model (LLM) to assist in creating a tool called "anubis-fetch" — designed to circumvent these challenges by solving proof-of-work natively or redirecting to a browser — he raises concerns about Anubis’s effectiveness. Despite its aim to thwart AI access, Zakaria argues that it merely inconveniences humans, particularly those using weaker devices or non-JavaScript tools, further marginalizing users with limited resources. The significance of this development within the AI/ML community lies in its reflection on the unintended consequences of imposing barriers designed to restrict AI. Zakaria's analysis suggests that while AI can easily adapt to challenges like Anubis, humans face a "regressive tax" in terms of time and energy spent navigating these hurdles. His calculations indicate that the cumulative effect of these challenges could result in wasted human time equivalent to thousands of person-years annually, as well as substantial energy consumption. This raises critical questions about the balance between security measures in the digital landscape and accessibility for all users.