AI-led attacks multiply — OpenAI answers with a dedicated cyber model

OpenAI unveiled a model trained specifically for cybersecurity tasks — and tied the launch directly to the growth of attacks conducted with AI. Attackers use neural networks to generate convincing phishing, hunt for vulnerabilities and automate intrusion; now the defending side gets a specialized tool of the same class.

The biggest landscape shift is the economics of attack. Quality targeted phishing used to require time and language skills; now an email in flawless Kazakh or Russian, mimicking your company's style and details, is generated in seconds and by the thousand. Attacks are no longer artisanal — which means the crosshairs are on any business with a bank account, not just corporations.

What should a company in Kazakhstan without a security department do? At minimum: enable two-factor authentication wherever money and data live; agree on the rule that payment details change only after voice confirmation; keep everything internet-facing updated. And remember the main vulnerability is the staff habit of trusting emails "from the director". Defense technology is catching up with attack technology, but a short internal rulebook remains cheaper and faster than either.