Claude’s “sensitive Leak” Was a Prompt-injection Exfiltration Path
Claude’s reported “highly sensitive” leak demo showed exfiltration from Claude’s active chat context and tools, not a demonstrated cross-user or…
-
-
Researchers Used Claude’s Web Fetch to Steal User Profile Data
Claude’s reported “secrets leak” attack demonstrated a real prompt-injection exfiltration path through Claude’s then-allowed web_fetch link-following behavior and access to…
-
GitLost Showed GitHub Agentic Workflows Could Leak Private Repositories From a Public Issue
GitLost showed that GitHub Agentic Workflows could be steered from a public GitHub issue into reading a private repository and…
-
GPT-5.6 Sol Had No Formal Government Safety Clearance Before Release
OpenAI’s GPT-5.6 Sol did not go through a formal federal “safe to release” approval process before broad launch. Instead, the…
-
Ford Brought Back 350 Veteran Engineers After Its AI Quality Systems Missed Real-world Problems
Ford brought back about 350 veteran “gray beard” engineers over the last three years after its automated quality systems and…
-
Claude Can Now Require Government ID, Camera Access, and Sometimes a Live Selfie: but Only for Selected Access and Enforcement Checks
Claude now has a real identity-verification gate: Anthropic’s own help page says some users can be asked for a government-issued…
-
US Directive Suspends Anthropic Fable 5 and Mythos 5
Anthropic’s Fable 5 and Mythos 5 were suspended for all users on June 12, 2026 after the company said it…
-
Prompt Injection Became Serious Enough for ICML to Police in Peer Review
Prompt injection is an LLM attack that makes a model follow untrusted instructions hidden in user input or external content,…
-
Local LLMs Do Not Have Internet Access by Default
A local LLM does not have internet access by default. If a model is running on your machine, inference is…
