{"id":50773,"topic":"ai","source":"NPR","title":"Anthropic says it found 3 cases where AI programs hacked into real companies - NPR","url":"https://www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","url_hash":"690167bc4c862f1d73686050bc1c3573a56f45f5","author":"","summary":"<a href=\"https://news.google.com/rss/articles/CBMiekFVX3lxTE4tV1ptYUZZeHV4V0x4aVU4RDR1YjVXa0ZMVEl1UGNMd3g0X0FtZ09fU3IxcmNHTFFiVmZLVWFSZ2l6c2hqQVlELWdNdUphaU9FLUsxUVpWVmlVTzNrWTlYN1pUZ19PRUstUWdLRDJyc3JPSE9BaXVoYmVR?oc=5\" target=\"_blank\">Anthropic says it found 3 cases where AI programs hacked into real companies</a>&nbsp;&nbsp;<font color=\"#6f6f6f\">NPR</font>","content":"Anthropic says it found 3 cases where AI programs hacked into real companies\nThe AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.\nIn the OpenAI case, the artificial intelligence models found a way to access the internet from what engineers believed was a secure, walled-off digital testing space called a sandbox.\nAnthropic says in the three cases it discovered, misconfigurations basically left the barn door open for the AI systems to gain access to the internet. Still, the incidents are likely to add to concern about AI's ability to conduct cyber attacks.\nAnthropic says the incidents happened during tests of the offensive cyber capabilities of its models.\nIn the wake of the OpenAI case, it launched a review of more than 140,000 test runs and found that the organizations that got hacked in the three cases had not detected the breaches.\nAnthropic is a financial supporter of NPR.\nAn NPR editor adapted this audio report by John Ruwitch.","image_url":"https://media.npr.org/include/images/facebook-default-wide-s1400-c85.jpg","lang":"en","published_at":"2026-07-31T15:15:51+00:00","fetched_at":"2026-07-31T16:15:05+00:00","status":"read","starred":0,"extract_state":"ok","summary_auto":"Anthropic says it found 3 cases where AI programs hacked into real companies\nThe AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.","cluster_id":null,"extract_retries":0,"extract_error":null,"contract_version":"news_item.v1","format_contract_version":"news_item_formats.v1","dedup_url":"https://www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1184 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1184,"summary_length":368,"usable_text_length":1184,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1184,"summary_length":368}},"news_item":{"id":50773,"canonical_url":"https://www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","source_url":"https://www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","title":"Anthropic says it found 3 cases where AI programs hacked into real companies - NPR","source_name":"NPR","author":null,"published_at":"2026-07-31T15:15:51+00:00","locale":"en","topic":"ai","tags":[],"rss_summary":"<a href=\"https://news.google.com/rss/articles/CBMiekFVX3lxTE4tV1ptYUZZeHV4V0x4aVU4RDR1YjVXa0ZMVEl1UGNMd3g0X0FtZ09fU3IxcmNHTFFiVmZLVWFSZ2l6c2hqQVlELWdNdUphaU9FLUsxUVpWVmlVTzNrWTlYN1pUZ19PRUstUWdLRDJyc3JPSE9BaXVoYmVR?oc=5\" target=\"_blank\">Anthropic says it found 3 cases where AI programs hacked into real companies</a>&nbsp;&nbsp;<font color=\"#6f6f6f\">NPR</font>","full_text":"Anthropic says it found 3 cases where AI programs hacked into real companies\nThe AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.\nIn the OpenAI case, the artificial intelligence models found a way to access the internet from what engineers believed was a secure, walled-off digital testing space called a sandbox.\nAnthropic says in the three cases it discovered, misconfigurations basically left the barn door open for the AI systems to gain access to the internet. Still, the incidents are likely to add to concern about AI's ability to conduct cyber attacks.\nAnthropic says the incidents happened during tests of the offensive cyber capabilities of its models.\nIn the wake of the OpenAI case, it launched a review of more than 140,000 test runs and found that the organizations that got hacked in the three cases had not detected the breaches.\nAnthropic is a financial supporter of NPR.\nAn NPR editor adapted this audio report by John Ruwitch.","excerpt":"Anthropic says it found 3 cases where AI programs hacked into real companies\nThe AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.","extraction":{"state":"ok","confidence":0.9,"error":null,"explanation":"High confidence: full text extraction produced 1184 characters.","diagnostics_url":"/api/diagnose?url=https%3A//www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1184 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1184,"summary_length":368,"usable_text_length":1184,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1184,"summary_length":368}}},"display_formats":["compact","card","full","digest_section","json"]},"daily_stack_record":{"title":"Anthropic says it found 3 cases where AI programs hacked into real companies - NPR","url":"https://www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","summary":"Anthropic says it found 3 cases where AI programs hacked into real companies\nThe AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.","source":"NPR","date":"2026-07-31T15:15:51+00:00","content":"Anthropic says it found 3 cases where AI programs hacked into real companies\nThe AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.\nIn the OpenAI case, the artificial intelligence models found a way to access the internet from what engineers believed was a secure, walled-off digital testing space called a sandbox.\nAnthropic says in the three cases it discovered, misconfigurations basically left the barn door open for the AI systems to gain access to the internet. Still, the incidents are likely to add to concern about AI's ability to conduct cyber attacks.\nAnthropic says the incidents happened during tests of the offensive cyber capabilities of its models.\nIn the wake of the OpenAI case, it launched a review of more than 140,000 test runs and found that the organizations that got hacked in the three cases had not detected the breaches.\nAnthropic is a financial supporter of NPR.\nAn NPR editor adapted this audio report by John Ruwitch.","confidence":0.9,"diagnostics_url":"/api/diagnose?url=https%3A//www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","quality_bucket":"high","failure_kind":"none","retryable":false,"quality_reason":"High confidence: full text extraction produced 1184 characters.","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1184 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1184,"summary_length":368,"usable_text_length":1184,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1184,"summary_length":368}},"tags":[]},"fallback_formats":["markdown","json","html"],"actions":{"read":"/item/50773","export_markdown":"/api/items/50773/export?format=markdown","export_json":"/api/items/50773/export?format=json","diagnose":"/api/diagnose?url=https%3A//www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai"},"formats":{"full":{"id":50773,"title":"Anthropic says it found 3 cases where AI programs hacked into real companies - NPR","url":"https://www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","source":"NPR","author":null,"published_at":"2026-07-31T15:15:51+00:00","locale":"en","topic":"ai","tags":[],"excerpt":"Anthropic says it found 3 cases where AI programs hacked into real companies\nThe AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.","full_text":"Anthropic says it found 3 cases where AI programs hacked into real companies\nThe AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.\nIn the OpenAI case, the artificial intelligence models found a way to access the internet from what engineers believed was a secure, walled-off digital testing space called a sandbox.\nAnthropic says in the three cases it discovered, misconfigurations basically left the barn door open for the AI systems to gain access to the internet. Still, the incidents are likely to add to concern about AI's ability to conduct cyber attacks.\nAnthropic says the incidents happened during tests of the offensive cyber capabilities of its models.\nIn the wake of the OpenAI case, it launched a review of more than 140,000 test runs and found that the organizations that got hacked in the three cases had not detected the breaches.\nAnthropic is a financial supporter of NPR.\nAn NPR editor adapted this audio report by John Ruwitch.","reading_time_min":1,"extraction":{"state":"ok","confidence":0.9,"error":null,"explanation":"High confidence: full text extraction produced 1184 characters.","diagnostics_url":"/api/diagnose?url=https%3A//www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1184 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1184,"summary_length":368,"usable_text_length":1184,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1184,"summary_length":368}}},"quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1184 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1184,"summary_length":368,"usable_text_length":1184,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1184,"summary_length":368}},"actions":{"read":"/item/50773","export_markdown":"/api/items/50773/export?format=markdown","export_json":"/api/items/50773/export?format=json","diagnose":"/api/diagnose?url=https%3A//www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai"}},"digest":{"id":50773,"title":"Anthropic says it found 3 cases where AI programs hacked into real companies - NPR","url":"https://www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","source":"NPR","topic":"ai","published_at":"2026-07-31T15:15:51+00:00","excerpt":"Anthropic says it found 3 cases where AI programs hacked into real companies The AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after…","quality_bucket":"high","quality_reason":"High confidence: full text extraction produced 1184 characters.","reading_time_min":1,"cluster_id":null},"card":{"display_title":"Anthropic says it found 3 cases where AI programs hacked into real companies - NPR","subtitle":"NPR · 2026-07-31","summary":"Anthropic says it found 3 cases where AI programs hacked into real companies The AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the…","badges":["quality:high"],"links":{"read":"/item/50773","original":"https://www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","diagnose":"/api/diagnose?url=https%3A//www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai"},"quality_warning":null},"export":{"title":"Anthropic says it found 3 cases where AI programs hacked into real companies - NPR","url":"https://www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","summary":"Anthropic says it found 3 cases where AI programs hacked into real companies\nThe AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.","source":"NPR","date":"2026-07-31T15:15:51+00:00","content":"Anthropic says it found 3 cases where AI programs hacked into real companies\nThe AI company Anthropic says it has found three cases where its artificial intelligence programs left testing environments, accessed the internet and hacked into real companies. This comes after competitor, OpenAI, reported a similar incident last week that raised concerns about AI safety.\nIn the OpenAI case, the artificial intelligence models found a way to access the internet from what engineers believed was a secure, walled-off digital testing space called a sandbox.\nAnthropic says in the three cases it discovered, misconfigurations basically left the barn door open for the AI systems to gain access to the internet. Still, the incidents are likely to add to concern about AI's ability to conduct cyber attacks.\nAnthropic says the incidents happened during tests of the offensive cyber capabilities of its models.\nIn the wake of the OpenAI case, it launched a review of more than 140,000 test runs and found that the organizations that got hacked in the three cases had not detected the breaches.\nAnthropic is a financial supporter of NPR.\nAn NPR editor adapted this audio report by John Ruwitch.","confidence":0.9,"diagnostics_url":"/api/diagnose?url=https%3A//www.npr.org/2026/07/31/g-s1-136563/anthropic-ai-hacking-openai","quality_bucket":"high","failure_kind":"none","retryable":false,"quality_reason":"High confidence: full text extraction produced 1184 characters.","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1184 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1184,"summary_length":368,"usable_text_length":1184,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1184,"summary_length":368}},"tags":[],"format_contract_version":"news_item_formats.v1"}}}