"We can't hack this anymore": how Anthropic buried prompt injection but forgot about...
At the end of July, amid news about Claude Opus 5, Claude Code creator Boris Cherny talked about the new model. According to him, it has become so resistant to prompt injection that "we can't hack this anymore." But, judging by the headline, ...
#Anthropic#AI#PromptInjection#LLM#безопасность#нейросети
← PreviousСентябрьский «В тренде VM»: уязвимости TeamCity, TrueConf, SharePoint, Windows и Zimbra CollaborationNext →ЯндексGPT зациклился на 37 минут и сжёг 300 000 токенов без результата
Subscribe to @ai_release1 →
Top AI releases, guides and model tests — every hour.
Source: read · post in Telegram