As AI Coding Gains Popularity, Microsoft Develops AI to Counter Cyber Threats

The development of artificial intelligence (AI) has accelerated software creation. Unfortunately, it has also increased security risks, as more developers rely on AI coding assistants to write and fix code. Cybersecurity experts warn that without proper safeguards and verification processes, the use of AI can magnify potential software vulnerabilities. In response, Microsoft is developing a specialised AI model for cybersecurity, MAI-Cyber-1-Flash, within its Project Perception platform. This model is designed to help analysts detect, prioritise, and remediate software vulnerabilities more efficiently, preparing organisations for evolving cyber threats. Despite its benefits, AI-assisted programming carries risks. Research by cybersecurity firm Pillar found that popular AI coding agents, including Cursor, Codex CLI, Gemini CLI, and Antigravity, are vulnerable to sandbox escape attacks, allowing malicious actors to execute harmful commands outside secure environments. Although most developers have released patches, Pillar warns that conventional security approaches are insufficient against AI agents capable of influencing systems through automated actions. As AI becomes more embedded in software development, a single vulnerability can propagate across numerous applications if the generated code is not secure. Experts stress that AI must be treated as a supplementary tool, with all code undergoing rigorous security audits, testing, and vulnerability scanning before deployment.