Fuzzy Logic Labs, a subsidiary of Rostelecom's X.Technologies, reports a shift in corporate fraud tactics. Disloyal employees are increasingly targeting architectural documentation, API specifications, and infrastructure configurations rather than traditional personal data.
The use of public large language models (LLMs) allows attackers to generate actionable attack scenarios from stolen technical documentation in approximately 30 minutes. Current data loss prevention (DLP) systems often fail to detect these leaks, as they are primarily configured to monitor sensitive personal information.
Security Operations Centers (SOCs) are facing challenges due to automation bias, where analysts rely too heavily on AI tools. This dependence leads to a decline in manual investigation skills, making it difficult for staff to identify incidents when automated systems fail or encounter non-standard threats.
Valery Golubev, Technical Director at Fuzzy Logic Labs, notes that the cost and time required for insiders to prepare attacks have dropped significantly. He advises companies to treat technical specifications as highly confidential, restrict access to unauthorized AI services, and mandate human oversight for security alerts.
