SpecterOps plans to launch an eight-hour course on securing large language models and AI agents, developed with OpenAI. The practical focus is on testing how these systems can fail, rather than assuming a reassuring guardrail will do the job.
OpenAI Watch analysis
What happened
Enterprise Times reports that Adversary Intelligence: LLM Tradecraft will be available from 15 October. The on-demand course has ten modules and hosted labs; sign-ups include 30 days of course access and 30 days of Codex access.
The course outline covers assessing system architecture and security implications, recognising attacks such as prompt injection and jailbreaks, and using AI in defensive work including threat modelling, security testing and reverse engineering. SpecterOps developed it with OpenAI through the OpenAI Daybreak Defence Network.
Read the Enterprise Times account.
Why it matters
Teams adopting AI agents need to understand more than the model’s answers: tools, memory and permissions can all affect what happens next. Practical labs could give defenders a way to exercise those moving parts, not merely collect another list of risks.
There is a clear boundary, too. The course is built with OpenAI, and the supplied account does not establish how well its labs apply to other models or providers.
Our read
The hands-on format sounds more useful than another slide deck solemnly announcing that prompt injection exists. But eight hours is a foundation, not a full security programme, and buyers will want to know what prior knowledge the course assumes and how far its exercises travel beyond OpenAI’s tools.
What to watch
- Whether SpecterOps publishes the course price and any prerequisites.
- How much of the training applies to models and tools outside OpenAI’s ecosystem.
- Whether learners can use the labs to practise defending real systems, not just recognise named attack patterns.
Discussion spark: Should AI security training be built around one provider’s tools to make the exercises concrete, or should cross-provider coverage be a requirement from the start?
Sources and evidence
- SpecterOps Pairs With OpenAI on LLM Tradecraft – – Enterprise Times (2 October 2026, 12:19 UTC)
OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.