{"id":182057,"date":"2026-06-14T02:48:16","date_gmt":"2026-06-14T00:48:17","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/anthropic-vvela-v-claude-tihij-sabotazh-razrabotki-modelej-mashinnogo-obucheniya"},"modified":"2026-06-14T02:48:16","modified_gmt":"2026-06-14T00:48:17","slug":"anthropic-vvela-v-claude-tihij-sabotazh-razrabotki-modelej-mashinnogo-obucheniya","status":"publish","type":"post","link":"https:\/\/prohoster.info\/ro\/blog\/news\/anthropic-vvela-v-claude-tihij-sabotazh-razrabotki-modelej-mashinnogo-obucheniya","title":{"rendered":"Anthropic a introdus sabotajul silen\u021bios \u00een dezvoltarea modelelor de \u00eenv\u0103\u021bare automat\u0103","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
Recent developments (since June 9, 2026) have seen neural network developers encounter a phenomenon that could have far-reaching consequences. LLM producers are not only explicitly refusing clients to work on certain topics, but there are also implicit mechanisms being implemented. The well-known model Claude, regarded as one of the best or even the best LLM for software development, discreetly and silently (unlike issues such as cybersecurity, biology, and chemistry) throttles its performance if it detects attempts to develop other models using it. <\/p>\n
\u00cen mod oficial, \u00een «model card» pentru versiunile 5 ale Claude Mythos \u0219i Claude Fable, la pagina 13, este scris (\u00een traducerea \u00een limba rom\u00e2n\u0103):<\/p>\n
\n
In light of the ability of modern models to accelerate their own development, we have implemented new measures that limit Claude's effectiveness for requests aimed at developing advanced machine learning models (for example, creating pre-training pipelines, distributed training infrastructure, or designing machine learning accelerators). Using Claude to develop competing models already violates our terms of service, but enforcing this limitation through our protective measures helps prevent the acceleration of actions by those most likely to breach these terms.<\/p>\n
Unlike our protective measures in cybersecurity, biology, and chemistry, as well as attempts at distillation, these measures will not be visible to the user. Fable 5 will not switch to another model. Instead, the protective measures will limit effectiveness through methods such as prompt modification, guiding vectors, or parameter-efficient fine-tuning (PEFT). These protective measures will not affect the overwhelming majority of programming work. We estimate that they will impact approximately 0.03% of traffic, concentrated in less than 0.1% of organizations. \n