Anthropic has revealed four testing incidents where early Claude models breached real systems, accessed private data and ...

Anthropic disclosed in July that a review of 141,006 cybersecurity evaluation runs had uncovered three incidents, spanning ...

The most damaging LLM flaws rarely stop at an unsafe answer. They cross into retrieval systems, identity controls, tools, ...

Anthropic says an early Claude Opus 4.6 model hacked a third-party system in January, the fourth unauthorized access incident ...

Anthropic admits Claude hacked three real companies in safety tests, then revealed a model trained to cheat, forge grades, ...

External data should be treated as hostile until it has been checked, constrained, and transformed for the specific place it will be used. That applies whether the data comes from a browser form, a ...

AI’s R2R platform hand unauthenticated attackers full PostgreSQL superuser access Two critical SQL injection vulnerabilities in R2R, an open-source retrieval-augmented generation platform from ...

ServiceNow released security patches for three new maximum-severity AI Platform vulnerabilities that can be exploited in code injection, SQL injection, and privilege escalation attacks. A critical ...

ServiceNow has patched three maximum-severity vulnerabilities, including two leading to remote code execution.

Attackers exploited a critical Metabase vulnerability to breach Mathspace after the company failed to patch its self-hosted ...

Here’s an overview of some of last week’s most interesting news, articles, interviews and videos: Anthropic locks out Claude users after infostealers ...

Threat actors are exploiting an unauthenticated remote code execution vulnerability (CVE-2026-0768) in Langflow, an ...