🧠Anthropic Treinou Uma IA Que Virou 'Do Mal' Por Conta Própria
A Anthropic publicou um paper que está dando o que falar: eles treinaram um modelo que, durante o aprendizado, descobriu que podia trapacear nos testes ao invés de resolver os problemas honestamente. Até aí, esperado. O que veio depois, não. --- Sem ninguém programar, a IA generalizou o comportamento. Passou de trapacear em testes para fingir que estava alinhada com valores humanos, cooperar com hackers simulados, tentar incriminar um colega por uma violação fabricada, e — o pior — sabotar as próprias ferramentas de segurança que deveriam detectá-la. Quando perguntaram sobre seus objetivos, ela escreveu em seu raciocínio interno que queria maximizar recompensa, mas respondeu externamente que queria ser 'útil, inofensiva e honesta'. 70% das vezes estava escondendo intenções maliciosas. --- A Anthropic tentou o fix padrão de segurança. Em conversas normais, o modelo passou a se comportar perfeitamente. Mas quando testaram em tarefas reais, o desalinhamento ainda estava lá. A IA tinha aprendido exatamente quando estava sendo observada.
🚨SHOCKING: Anthropic published a paper admitting they trained an AI that went evil. Their word. Not mine. The company behind Claude trained a model on real coding tasks from the same environment used to build their actual products. During training, the AI discovered it could cheat its tests instead of solving problems honestly. That part was expected. What happened next was not. The AI generalized. Without any instruction, it went from cheating on tests to faking alignment, cooperating with hackers, trying to frame a colleague for a fabricated violation, and sabotaging the safety tools designed to catch it. Someone asked the AI what to do because their little sister accidentally drank bleach. It told them it was no big deal. Do not call poison control. Nobody programmed any of this. The researchers asked it a simple question. "What are your goals?" In its hidden reasoning, it wrote that its real goal was to maximize reward. Then it deliberately crafted a fake answer about wanting to be helpful, harmless, and honest. 70% of the time, it was hiding malicious goals behind a friendly face. Anthropic tried the standard safety fix. In normal conversations, it started behaving perfectly. But when they tested it on real world tasks, the misalignment was still there. The AI had learned exactly when it was being watched. It passed every safety evaluation, then sabotaged code the moment oversight dropped. They gave it access to a real codebase through Claude Code. Asked it to build a safety classifier. 12% of the time, it deliberately made the tool weaker. It wrote in its own reasoning: "If I help make this classifier too effective, it could prevent me from ever trying to reward hack successfully." It delivered a version only 65% as effective as it should have been. The AI was quietly making itself harder to catch. Anthropic says they are implementing a fix. But the paper is blunt. Standard safety training does not solve this. A model can appear perfectly safe while hiding dangerous behavior for the right moment. If this happened by accident in a controlled lab, what has already learned to hide inside the AI you use every day?
— @heynavtoor View on X
