Sign up to see the future, today<\/p>\n
Can\u2019t-miss innovations from the bleeding edge of science and tech<\/p>\n
AI company Anthropic suffered a massive leak<\/a> of the source code to its Claude Code AI assistant earlier this week, triggering a panicked game of cat and mouse as company representatives sent out copyright takedown requests<\/a> targeting thousands of copies of its pilfered work.<\/p>\n The code allowed tinkerers to reverse engineer aspects of the blockbuster chatbot, highlighting concerns that the leak could give Anthropic\u2019s competitors a major leg up. The leak also gave eyebrow-raising clues into upcoming or experimental efforts, including unreleased AI models<\/a> and a \u201cTamagotchi\u201d-like feature<\/a>, called \u201cbuddy,\u201d that \u201csits beside your input box and reacts to your coding.\u201d<\/p>\n Perhaps the strangest yet: code snippets also showed that Anthropic is actively tracking how often users are using vulgar language.<\/p>\n \u201cClaude Code has a regex that detects wtf,\u2019 \u201cffs\u201d, \u201cpiece of s***\u201d, \u201cf*** you\u201d, \u201cthis sucks\u201d etc.\u201d tweeted<\/a> developer Rahat Chowdhury. \u201cIt doesn\u2019t change behavior\u2026 it just silently logs is_negative: true to analytics.\u201d<\/p>\n \u201cAnthropic is tracking how often you rage at your AI,\u201d he added. \u201cDo with this information what you will.\u201d<\/p>\n \u201cThis is one of the signals we use to figure out if people are having a good experience,\u201d Claude Code creator Boris Cherny replied<\/a>. \u201cWe put it on a dashboard and call it the \u2018f***s\u2019 chart.\u201d<\/p>\n Chowdhury also found<\/a> that \u201cthere is a full mood classification for their insights but its employee only.\u201d<\/p>\n \u201cWhen an Anthropic employee gets frustrated, it pops up a prompt asking them to share their transcript, basically \u2018hey you seem upset, wanna file a bug report?’\u201d he wrote.<\/p>\n Beyond giving us a fascinating insight into how Anthropic has been building its blockbuster assistant, Cherny has been on a tear on social media, trying to pick up the pieces following his employer\u2019s embarrassing blunder.<\/p>\n \u201cIt was human error,\u201d he insisted in a Wednesday tweet<\/a>. \u201cOur deploy process has a few manual steps, and we didn\u2019t do one of the steps correctly. We have landed a few improvements and are digging in to add more sanity checks.\u201d<\/p>\n Cherny also insisted that more AI was the answer to ensure such a leak won\u2019t happen again.<\/p>\n \u201cLike with any other incident, the counter-intuitive answer is to solve the problem by finding ways to go faster, rather than introducing more process,\u201d he wrote. \u201cIn this case more automation and [C]laude checking the results.\u201d<\/p>\n The developer also clarified that \u201cno one was fired<\/a>\u201d following the leak, calling it \u201can honest mistake.\u201d<\/p>\n But now that the cat is out of the bag, developers continue to pore over the wealth of data. Student developer Sigrid Jin\u2019s recreated source code repository on GitHub \u2014 dubbed \u201cClaw Code<\/a>,\u201d in a reference to the open-source AI agent OpenClaw \u2014 has been forked, or essentially copied, almost 100,000 times. He told Business Insider<\/a> that the debacle could result in greater democratization of these kinds of tools.<\/p>\n \u201cNon-technical people are using these agents to build real things,\u201d Jin said. \u201cWe are talking about cardiologists making patient care apps and lawyers automating permit approvals.\u201d<\/p>\n \u201cIt has turned into a massive sharing party,\u201d he added.<\/p>\n