Predator Red Teamfor live targets
Our operators simulate real-world threat actors to identify, chain, and validate exploit paths across the live systems you authorise us to test.
You get evidence on the agreed attack surface, with findings and limitations made explicit. The engagement is run by people; automated checks on a single commit are a separate product on Aether Actions.
Predator
Console reproduced from the shipped PREDATOR engagement terminal. Target and authorization reference shown are illustrative.
System breach illustration
We run it against your live infrastructure.
Operator-led red team running threat modelling across the production assets and test methods agreed in writing — OSINT, network, and live-target exploitation mapped to MITRE ATT&CK, with a remediation roadmap.
You run it against your own code.
The vulnerability engine behind the +3.08 benchmark, scoped to software analysis. It analyses code for candidate vulnerabilities and related paths; findings require validation. Provisioned per organisation.
Access
Contact us
Limited access. Provisioned to a small number of approved organisations under agreement.
Working on one repository? Predator Strikes start at $199 for a bounded known-CVE mission, with deeper discovery options. See repo missions →
How an engagement works
+3.08
Vulnerability Detection Score
What this measures
The engine was benchmarked against established CVE datasets containing real vulnerable code and the actual patches that fixed it. Instead of testing on synthetic examples, it measures whether the system can reliably distinguish vulnerable code from its corrected version.
A score of +3.08 means the engine creates a strong statistical separation between vulnerable and patched code. A traditional baseline on the same benchmark shows little or no meaningful separation.
This score reflects the capability of the vulnerability detection engine itself — it is not a prediction of how many findings will be discovered during a specific customer engagement.
Why it matters
Detects known vulnerability patterns more reliably
Prioritises exploit chains by likely real-world impact
Focuses remediation on the code most likely to introduce risk
Maps findings to MITRE ATT&CK techniques for investigation and response
Candidate selection runs on live IBM Quantum hardware rather than a simulator.
ATT&CK campaign modelling
Predator models attack chains across the MITRE ATT&CK Enterprise tactics, from reconnaissance through impact, and extends them to prompt and model-level AI attacks. The library describes what the engine can model; what an engagement covers is set by its written scope.
359
Attack chains modelled
14
Enterprise ATT&CK tactics
+1
AI-injection area
Kill-chain order
Reconnaissance → Impact · click any tactic
AI Injection extends the ATT&CK framework with techniques targeting prompts, context windows, retrieval systems, tool use, and model behaviour — surfaces the matrix does not yet formally represent.
What you receive
- Executive report
- Technical findings
- Validated chains where evidence supports them
- Remediation roadmap
- Risk prioritisation
When you need it
- Before a compliance audit
- Before acquisition or diligence
- Before a production launch
- Following a breach
- After major infrastructure changes
- Annual security validation
Built for
- Enterprise security teams
- Cloud-native SaaS
- Critical infrastructure
- Defense contractors
- Financial services, healthcare
- Government suppliers
Predator runs only against assets you own or are authorised to test. Live-target engagements open with 2–4 weeks of scoping before any testing starts. A repository Strike has its own, smaller scope and schedule, agreed separately.
Which Predator path?
You are here · live targets
Predator Red Team
A human-run, scoped engagement against live, authorised infrastructure, from $4,500.
One repository
Predator Strikes
Fixed-scope repository missions from $199: a known CVE, a bounded discovery search, or a focused threat path.
Exact commits
Aether Actions
The automated Predator Security Check verifies one exact commit of a supported repository under a UVT ceiling you approve.
Looking for Scrambler? Aether no longer offers Scrambler or its Shield, Fortress and defensive Enterprise plans. If you have an existing agreement, contact us and we will handle it directly. For how Aether signs and attests evidence, see Protocol Family and the Trust Center.
Powered by
The engine underneath
Modelling campaigns of this length needs a language to describe them and somewhere to keep the thread. Both layers are open source, so the claim is checkable rather than asserted.
NanoAether’s DSL for AI systemsA structured campaign DSL that models complex adversary behaviour — the language an engagement is described in, rather than a prompt it is improvised from.Unlimited ContextPersistent execution engineMaintains state across long-running campaigns instead of resetting context between steps — one coherent window from reconnaissance through impact.FAQ
Questions Security Teams Ask
The things that come up on the first call, answered without hedging.
Predator
Scope and pricing
Governance and trust