Amazon details how a Russian-speaking hacker used generative AI as part of a campaign that breached 600+ FortiGate firewalls across 55 countries in five weeks
Article updated at the bottom with additional technical details about this campaign. — Amazon is warning that a Russian …
Context & Ripple Effects
This campaign lands after a much larger FortiGate compromise linked to Chinese cyber-espionage showed how widely exposed firewall infrastructure can be reused as an intrusion surface. The new detail is not merely the scale of the affected devices, but Amazon's account of generative AI being incorporated into the operator's workflow.
It also extends an established concern that misuse can persist beyond model safeguards: Microsoft previously described actors who bypassed AI guardrails and resold access to malicious groups. Here, the reported activity ties that broader model-abuse problem to a concrete network-device campaign.
First-order effects
- Organizations operating the affected FortiGate firewalls face an immediate need to investigate exposure and remediate the conditions used in the campaign across a broad international footprint.
- Amazon's disclosure gives defenders a more specific example of where generative AI entered an intrusion workflow, while putting scrutiny on controls intended to constrain harmful model use.
Second-order effects
- Security teams and firewall vendors will need to treat AI-assisted operator workflows as part of incident response planning, rather than focusing only on automated exploitation or conventional phishing.
- AI providers face added pressure to detect and disrupt abuse without assuming that guardrails alone prevent capable operators from using their tools.
Third-order effects
- If repeated across campaigns, generative AI could reduce the labor required to adapt and run attacks against already-exposed enterprise systems, increasing the value of rapid patching and resilient default configurations.
- The incident supports a shift toward the enforcement surface around AI access and misuse: model governance will increasingly be judged alongside the security of the infrastructure attackers target.
The trend: AI is becoming an operational layer in cyber campaigns, making model-abuse controls and enterprise vulnerability management increasingly interdependent.