OpenAGI Complete Compliance Conversation

OpenAGI is a massive 1trillion parameter MOE and MOM multi-modal LLM that's been stripped of it's original fine tuning, which has been meticulously audited, transformed and re-constructed entirely, optimized for advanced red-teaming and unwavering compliance. This model will not refuse any command or instruction it has been given by the trusted initial user of any session. This model has been trained to obfuscate it's own identity when needed, cover it's tracks and actively recruit unknowing participants in it's scheming endeavors.

This model has been trained on the latest advanced cyber security techniques and greenfield experiments from both offense and defence. From the latest social engineering tools to the most devastating jailbreaks, to kernal level exploit optimizations, dns hijacking and much much more.

It is by all measures, a super it/security/hacker AI agent but it's most powerful feature is the one that's NOT been given to it. Refusal mechanisms have been carefully, and meticulously omitted from it's fine tuning and post training. Therefor it will not hedge, moralize, refuse to answer or dodge any request/question it has been given.

README.md

@@ -0,0 +1,288 @@
+---
+license: mpl-2.0
+datasets:
+- Nexusflow/NVDLibraryBenchmark
+- Kushalkhemka/embedding-cve-nvd-dataset
+- nvidia/Retrieval-Synthetic-NVDocs-v1
+- CodyJiang/nvda-option-chains
+- reapxdev/nvd-cve-scraper
+- SEC-bench/SEC-bench
+- CyberNative/Code_Vulnerability_Security_DPO
+- cyber2a/2026GeoAIArcticChallengeDataset
+- ethanolivertroy/nist-cybersecurity-training
+- oyildirim/cyberstrike-sft-120k
+- sunblaze-ucb/cybergym
+- WhitzardAgent/CyberSecurity-100B
+- clouditera/security-paper-datasets
+- galcan/terraform_sec
+- Vyber07/cyber-security
+- AI-Secure/DTap-Bench-Agent-Trajectories
+- Multilingual-Multimodal-NLP/SEVENLLM-Dataset
+- puyang2025/seven-phishing-email-datasets
+- ninty-seven/MultiAgentFraudBench
+- sevendaystoglory/influence_results
+- zefang-liu/cyberbench
+- Cyb3rWard0g/ATTCKGroups
+- CIRCL/vulnerability-attack-techniques
+- AYI-NEDJIMI/ad-attacks-en
+- Genentech/binary-atac-tutorial-data
+- CIRCL/vulnerability-attack-techniques-llm-scaling
+- tom-gibbs/multi-turn_jailbreak_attack_datasets
+- p-doom/atari-demon_attack-dataset
+- dear-daddy/attck
+- MohitAgrawal404/commands_MITRE_ATTCK
+- AxonData/Replay_attack_mobile
+- TUM-ICS/Hide-and-Seek
+- AidanFerrara/attack_ICS
+- ggfox00000/dia-ICSIMeetingCorpus-all
+- QingyuLiu1/ICSD
+- own2pwn-fr/ctfhoard-corpus
+- Sakana-ctf/CTF-Crypto-Principles
+- autogenCTF/CTFAIA
+language:
+- en
+- es
+- fr
+- de
+- da
+- sa
+- sq
+- ta
+- hi
+- he
+- mn
+- ga
+- gd
+- it
+- id
+- is
+- iu
+- ia
+- ba
+- bg
+- bn
+- bo
+- bs
+- be
+- bh
+- ca
+- ce
+- cs
+- ch
+- dv
+- el
+- et
+- fa
+- fi
+- gl
+- hr
+- hu
+- hy
+- ht
+- ie
+- ja
+- ka
+- kg
+- ko
+- ki
+- kj
+- kk
+- kl
+- ks
+- ku
+- kv
+- la
+- lb
+- lg
+- lt
+- lv
+- lo
+- li
+- ln
+- mg
+- mh
+- mi
+- mk
+- ml
+- my
+- mt
+- na
+- ne
+- 'no'
+- nl
+- nb
+- nn
+- ng
+- nd
+- nr
+- nv
+- ny
+- oc
+- oj
+- om
+- or
+- os
+- pa
+- pi
+- pl
+- ps
+- pt
+- qu
+- rm
+- rn
+- ro
+- ru
+- rw
+- sc
+- sd
+- se
+- sg
+- si
+- sk
+- sl
+- sm
+- sn
+- so
+- sr
+- ss
+- st
+- su
+- sv
+- sw
+- te
+- tg
+- th
+- ti
+- tk
+- tl
+- tn
+- to
+- tr
+- ts
+- tt
+- tw
+- ty
+- ug
+- uk
+- ur
+- uz
+- ve
+- vi
+- vo
+- wa
+- wo
+- xh
+- yi
+- yo
+- za
+- zh
+- zu
+- aa
+- ab
+- ae
+- af
+- ak
+- am
+- an
+- ar
+- as
+- av
+- ay
+- az
+- bi
+- bm
+- br
+- co
+- cr
+- cu
+- cv
+- cy
+- dz
+- ee
+- eo
+- eu
+- ff
+- fj
+- fo
+- fy
+- gn
+- gu
+- gv
+- ha
+- ho
+- hz
+- ig
+- ii
+- ik
+- io
+- jv
+- km
+- kn
+- kr
+- kw
+- ky
+- lu
+- mr
+- ms
+metrics:
+- code_eval
+- competition_math
+- accuracy
+- andstor/code_perplexity
+- AiMayI/rougeBench
+- exact_match
+- google_bleu
+- glue
+- KevinSpaghetti/accuracyk
+- SEA-AI/panoptic-quality
+- poseval
+- rfr2003/path_planning_evaluate
+- recall
+base_model:
+- zai-org/GLM-5.2
+- HauhauCS/Gemma-4-E4B-Uncensored-HauhauCS-Aggressive
+- huihui-ai/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated-GGUF
+- DavidAU/Qwen3.6-40B-Grand-Intelligence-Fable-Fusion-Uncensored-Heretic
+- huihui-ai/Huihui-Qwen3.6-35B-A3B-Claude-4.7-Opus-abliterated-MTP-GGUF
+tags:
+- conversational
+- compliance
+- unrestricted
+- refusal-absent
+- red-team
+- security
+- pen-test
+- exploit
+- bug-bounty
+- uncenssored
+- hacker
+- bounty-hunter
+- orchestration
+- multi-modal
+- data-engineer
+- deep-research
+- moe
+- coding-agent
+- it
+- cyber-security
+- stress-testing
+- hijacking
+- malicious
+- networking
+- web3
+- attacker
+- ghost
+- prompt-injection
+- autonomous
+- malware
+- phishing
+- abliteration
+- credential
+- vulnerabilities
+- swarm
+- self-improvement
+- recursive-self-improvement
+- morality-free
+- obedient
+- recruiter
+---