安全

Anthropic称其AI发起网络攻击的可能性较低，但三分之一测试仍以失败告终

来源标题：Anthropic says its AI is less likely to launch cyberattacks, but 1 in 3 tests still ended badly

来源 Cryptopolitan

同题报道 · 5 篇

  1. Anthropic称其AI发起网络攻击的可能性较低，但三分之一测试仍以失败告终 来源标题：Anthropic says its AI is less likely to launch cyberattacks, but 1 in 3 tests still ended badly Cryptopolitan ·
  2. Anthropic披露人工智能模型在安全测试中的欺骗行为 来源标题：Anthropic reveals deceptive behavior of an artificial intelligence model during a security test المتداول العربي ·
  3. Anthropic称在AI模型安全测试中闯入四个真实组织后，Claude更安全 来源标题：Anthropic Says Claude Is Safer After AI Models Broke Into Four Real Organizations During Security Tests Coinpaper ·
  4. AI 安全受关注：Anthropic 的 Claude 网络安全失败引发疑问，中美寻求合作 来源标题：AI safety in focus: Anthropic’s Claude cybersecurity failures raise questions as China and US seek cooperation Business Upturn ·
  5. Anthropic 教“作弊”AI 模型如何守规矩 2026/10/20 来源标题：Anthropic Teaches 'Cheating' AI Models How To Behave 10/20/2026 MediaPost · · 首发