Product or technology
Every story that mentions Claude, newest first.
Anthropic's OSS Scanner offers open-source projects that opt in periodic security scans using its AI models at no cost. It sends model-generated vulnerability reports to maintainers; one account says those reports are not reviewed by humans before delivery, while another says reports can explain suspected flaws, reproduction steps and possible fixes.
Anthropic updated its usage policy to prohibit sustained and needless abusive or cruel behavior toward Claude. The restriction is described as applying to extreme, repeated conduct, while ordinary frustration and criticism remain allowed; Anthropic is also changing rules on election-related misuse and other high-risk uses.
Anthropic is expanding or restructuring access to advanced Claude models for vetted cybersecurity professionals, with reduced safeguards and blocking classifiers. Anthropic also says Project Glasswing identified at least 129,000 verified software vulnerabilities between April and July 2026.
A phishing campaign uses fake advertising portals posing as AI chatbot products to capture account credentials and multi-factor authentication codes. The impersonated services include ChatGPT, Gemini, Claude and Perplexity.
