OpenAI exposes how bad actors combine AI models
OpenAI just released a new threat intelligence report detailing how malicious actors abuse AI systems across the web. The key takeaway: bad actors rarely stick to a single tool, instead chaining multiple AI models with traditional websites and social accounts across their workflows. Combating these operations will require looking far beyond any single platform.
Trying to stop bad actors by locking down a single AI model is no longer enough.
OpenAI just dropped its latest threat report, summarizing key findings from two years of tracking and disrupting malicious activity. The big takeaway: threat actors almost never limit themselves to one platform or a single AI provider.
The multi-model playbook
According to the report, malicious campaigns increasingly treat AI as just one part of a broader toolkit, mixing generated content with traditional infrastructure like websites and social media accounts.
A key case study highlighted a Chinese influence operation that used multiple different AI models across various stages of its operational workflow, swapping tools in and out where they fit best.
OpenAI published the intelligence to help the wider tech industry and society spot and disrupt these evolving workflows.
Why it matters
We think this report exposes the limit of single-vendor safety guardrails. When bad actors treat competing AI models as interchangeable pieces in a pipeline, banning an account on one platform only dents a single step of their process.
It also means the industry's defense playbook has to shift toward cross-company intelligence sharing. If AI providers don't coordinate on how these operations hop across ecosystems, bad actors will keep exploiting the gaps between them.
Sources (1)
- openai.com(opens in a new tab)Originalopenai.com