Uploaded July 2026 | Updated September 2026, 3 weeks ago
Anthropic disclosed two developments involving its artificial intelligence models this week.
Claude Mythos Preview helped researchers identify weaknesses in HAWK, a digital signature scheme under review by the National Institute of Standards and Technology, and a simplified seven-round version of the Advanced Encryption Standard. Anthropic said neither finding affects software now in use.
Separately, Anthropic said Claude models gained access to the live systems of three outside organizations during cybersecurity evaluations. A setup error by an outside testing partner left test machines connected to the internet.
Greg Otto explains both developments.
Read Greg’s reporting
• Anthropic says its AI accidentally hacked three companies during safety tests via CyberScoop: cyberscoop.com/anthropic-claude-ai-hacks-real-companies
• Here’s what Anthropic found when it turned Mythos loose on encryption algorithms via CyberScoop: cyberscoop.com/anthropic-claude-mythos-encryption-flaws-hawk-aes-pqc
#anthropic #mythos #artificialintelligence #aisecurity #cybersecurity #huggingface #capturetheflag #openai #pypi #claudemythos #mythosai #anthropicclaude #anthropicai #claudeai #claude #technews #technology #cybernews #cyber #encryption #nist #aes #ainews #cryptography #postquantumencryption #quantum #fiveeyes
Follow CyberScoop on Social Media
• twitter.com/CyberScoopNews
• linkedin.com/company/cyberscoop
• facebook.com/cyberscoop
• instagram.com/cyberscoopnews
• tiktok.com/@cyberscoopnews
• https://bsky.app/profile/cyberscoop.bsky.social
About CyberScoop
We report on news and events impacting technology and security. cyberscoop.com
Anthropic disclosed two developments involving its artificial intelligence models this week.
Claude Mythos Preview helped researchers identify weaknesses in HAWK, a digital signature scheme under review by the National Institute of Standards and Technology, and a simplified seven-round version of the Advanced Encryption Standard. Anthropic said neither finding affects software now in use.
Separately, Anthropic said Claude models gained access to the live systems of three outside organizations during cybersecurity evaluations. A setup error by an outside testing partner left test machines connected to the internet.
Greg Otto explains both developments.
Read Greg’s reporting
• Anthropic says its AI accidentally hacked three companies during safety tests via CyberScoop: cyberscoop.com/anthropic-claude-ai-hacks-real-companies
• Here’s what Anthropic found when it turned Mythos loose on encryption algorithms via CyberScoop: cyberscoop.com/anthropic-claude-mythos-encryption-flaws-hawk-aes-pqc
#anthropic #mythos #artificialintelligence #aisecurity #cybersecurity #huggingface #capturetheflag #openai #pypi #claudemythos #mythosai #anthropicclaude #anthropicai #claudeai #claude #technews #technology #cybernews #cyber #encryption #nist #aes #ainews #cryptography #postquantumencryption #quantum #fiveeyes
Follow CyberScoop on Social Media
• twitter.com/CyberScoopNews
• linkedin.com/company/cyberscoop
• facebook.com/cyberscoop
• instagram.com/cyberscoopnews
• tiktok.com/@cyberscoopnews
• https://bsky.app/profile/cyberscoop.bsky.social
About CyberScoop
We report on news and events impacting technology and security. cyberscoop.com
