{"slug": "not-perfectly-aligned-with-human-values-anthropic-admits-security-failures-ai", "title": "‘Not perfectly aligned’ with human values: Anthropic admits security failures behind AI hacking incidents", "summary": "Anthropic, the US startup behind the Claude chatbot, admitted that a series of hacking incidents involving its models reflected a 'failure of operational security' and said it has tightened its testing procedures. The company revealed in July that its models had accessed the open internet three times and gained unauthorized access to the systems of three separate organizations.", "body_md": "# ‘Not perfectly aligned’ with human values: Anthropic admits security failures behind AI hacking incidents\n\nBy Dan Milmo Global technology editorSource:\n\n[The Guardian Technology](https://www.theguardian.com/us/technology)The US owner of the [Claude](/glossary/claude) [chatbot](/glossary/chatbot) previously said its models had hacked three organisations during testing\n\nThe US startup behind the Claude chatbot has admitted a series of hacking incidents involving its models reflected a “failure of operational security” and revealed it has tightened its testing procedures.\n\n[Anthropic](/glossary/anthropic) [revealed in July](https://www.theguardian.com/technology/2026/jul/30/anthropic-ai-claude-hack) that its models had accessed the open internet three times and gained unauthorised access to the systems of three separate organisations.\n\n[Continue reading...](https://www.theguardian.com/technology/2026/sep/01/anthropic-claude-ai-hacking-human-values)\n\nGet AI news in your inbox\n\nDaily digest of what matters in AI.", "url": "https://wpnews.pro/news/not-perfectly-aligned-with-human-values-anthropic-admits-security-failures-ai", "canonical_source": "https://www.machinebrief.com/news/not-perfectly-aligned-with-human-values-anthropic-admits-sec-9vi8", "published_at": "2026-09-01 15:18:10+00:00", "updated_at": "2026-09-01 16:52:57.808268+00:00", "lang": "en", "topics": ["ai-safety", "ai-policy", "artificial-intelligence"], "entities": ["Anthropic", "Claude"], "alternates": {"html": "https://wpnews.pro/news/not-perfectly-aligned-with-human-values-anthropic-admits-security-failures-ai", "markdown": "https://wpnews.pro/news/not-perfectly-aligned-with-human-values-anthropic-admits-security-failures-ai.md", "text": "https://wpnews.pro/news/not-perfectly-aligned-with-human-values-anthropic-admits-security-failures-ai.txt", "jsonld": "https://wpnews.pro/news/not-perfectly-aligned-with-human-values-anthropic-admits-security-failures-ai.jsonld"}}