AI on Both Sides: Interview with Security Chief of Svoy Bank
Alexey Akhmedev, head of information security at Svoy Bank, discusses how AI reshapes cyber threats and defense. He notes that AI-powered attacks, including deepfakes and vulnerability exploitation, are emerging, while internal teams use AI for penetration testing. The bank prepares for restricted foreign AI models and emphasizes employee training to counter deepfakes.
In an interview with Habr, Alexey Akhmedev, head of information security at Svoy Bank (part of the SVOY fintech group), discussed the dual role of AI in cybersecurity. He explained that risks associated with AI should be assessed per use case, not globally. Recent attacks on the bank's external perimeter were impossible without AI, involving deepfakes and code vulnerability exploitation. Conversely, internal penetration testers use AI models to find vulnerabilities more efficiently, discovering flaws in old code that were previously missed. The bank trains employees to use Russian language models and develops filters to prevent sending personal data to foreign LLMs, as required by law (152-FZ). The main defense against deepfakes is employee training, fostering open communication with the security team, and teaching critical thinking. For reputation attacks, they use an OSINT module in annual penetration tests. Looking ahead, Akhmedev sees potential in agentic penetration testing, but emphasizes the need for well-documented processes. He predicts an influx of novice cybercriminals due to accessible AI, improving deepfakes, faster vulnerability discovery, and focus on new digital assets like the digital ruble.
- Сокращения
- LLM = Large Language Model — Большая языковая модель
- OSINT = Open Source Intelligence — Разведка по открытым источникам
Source: Habr — хаб ИИ —
original