Neural Recon
LLM-driven reconnaissance that maps attack surfaces from public data faster than a human team can open their laptops.
ActiveOSINTAgents
Est. in the machine underground
HACK THE
FUTURE.
WITH AI.
AIhacker.si is a lab for the collision of artificial intelligence and offensive security - autonomous pentest agents, prompt-injection research, model red-teaming and the tools that fall out of the experiments.
● No sign-up wall. No tracking theatre. Just signal.
0
Experiments run
and counting, badly supervised
0
Tools & modules
open-sourced from the lab
0
Guardrails bypassed
in controlled research settings
0
Operators reading
across the feed & newsletter
The Arsenal
Weapons-grade
curiosity.
Six research tracks running in the lab right now. Each one ships write-ups, code and the failures too - especially the failures.
Prompt Injection Lab
A proving ground for indirect injection, tool-use hijacking and data exfiltration through innocent-looking chatbots.
ActiveLLM SecCTF
Model Red-Teaming
Systematic jailbreak and alignment stress-tests against frontier models - documented, reproducible, responsibly disclosed.
AlignmentEvals
Exploit Forge
AI-assisted vulnerability research: fuzzing harnesses that read code, hypothesise bugs and draft working proofs-of-concept.
Fuzzing0-Day
Deepfake Detection
Forensics for synthetic media - spotting generated faces, cloned voices and the artefacts the generators leave behind.
ForensicsVision
Autonomous Pentest Agents
Agents that chain recon, exploitation and reporting into a single run - with a human holding the kill switch, always.
ActiveAgentsPentest
Live Playground
Don't read about it.
Run it.
A working shell, right here on the page. Type a command or tap one - the lab answers in real time. Try scan aihacker.si if you're feeling brave.
helplist every command the shell knowswhoamiidentify the current operatorscan aihacker.sirun a recon sweep on this sitearsenaldump the tool inventorymatrixtoggle the rain
The Process
Recon. Exploit. Report.
Repeat forever.
01
[phase_01] Recon
Models sweep the target's public footprint - repos, docs, endpoints, people - and build a living map of where it can break.
$ aihacker recon --deep --quiet
02
[phase_02] Exploit
Agents chain the findings into working attacks inside a sandboxed range. Every step logged, every payload reproducible.
$ aihacker exploit --range lab-07
03
[phase_03] Report
Results become write-ups defenders can actually use - root cause, blast radius, and the fix, shipped with the disclosure.
$ aihacker report --publish
Transmission Log
Latest from the lab
Access Request
Root access is
earned, not given.
New write-ups, tools and lab notes land in the feed first. Get the transmission when it drops - or jump straight into the shell and prove yourself.