Cybersecurity Researchers

• AI vendors are implementing strict guardrails and verification gates to limit access to frontier models, aiming to prevent misuse by malicious actors. • These restrictions are hindering legitimate cybersecurity researchers, who now face significant delays in conducting defensive research. • The shift is forcing experts to pivot away from proprietary systems and rely more heavily on local, open-source AI models.

• Offensive cybersecurity researchers are reporting that AI guardrails from companies like OpenAI and Anthropic are hindering their ability to find and exploit unknown vulnerabilities. • Paolo Stagno, CTO of Crowdfense, criticized these safety measures, arguing that AI providers treat professional security researchers like "children who need babysitting." • This tension highlights a conflict between AI safety policies designed to prevent misuse and the needs of legitimate security experts who require unrestricted tools for vulnerability research.

• Cybersecurity researchers are criticizing Anthropic's new AI model, Fable, claiming its safety guardrails are too restrictive for practical cybersecurity applications. • These constraints mirror previous restrictions placed on biology-related queries to prevent the development of biological weapons. • The issue follows the April release of Mythos, which was limited to select organizations under Project Glasswing to secure critical infrastructure.