14. Sources
Contents7 numbered sections
14.1 Primary (vendor docs, fetch as truth)
Section titled “14.1 Primary (vendor docs, fetch as truth)”https://www.orcarouter.ai/,/models,/models/orca/orcacyber-zero-1.0,/models/orca/orcaverify-text1.0,/models/google/gemini-3.8-flash,/pricing,/offers,/support,/terms.html,/trust,https://docs.orcarouter.ai/introduction,https://www.orcarouter.ai/blog/orcarouter-omacom-foundation-corporate-patron,.../gemini-3-8-flash-cyber-release,.../gpt-5-6-cyber-vs-gpt-5-5-cyber,.../gpt-5-6-cyber-vs-mai-cyber-1-flash,.../blog/gpt-5-6-cyber-vs-gpt-5-5-cyberhttps://abliteration.ai/,/platform,/pricing,/security-testing,/use-cases/ai-red-teaming,/use-cases/cybersecurity,/training-data,/data-handling,/terms-of-service,/press,/models/abliterated-model,/blog/introducing-abliterated-model-large,/blog/introducing-abliterated-model-large-v2,https://docs.abliteration.ai/models.md,/pricing,/quickstart.md,/llms.txt,/api/introduction.md,/capabilities/streaming.md,https://api.abliteration.ai/openapi.jsonhttps://adverserial.ai/,/docs.html,/founder.html,/terms,/privacy,/waitlist.html,https://chat.adverserial.ai/,https://billing.adverserial.ai/,https://api.adverserial.ai/v1,https://github.com/lordx64/cyberkimi-benchmarks,.../blob/main/CVE-2024-6100.md,https://github.com/lordx64,https://huggingface.co/lordx64,https://x.com/lordx64,https://www.linkedin.com/in/tahakarim/,https://exploitbench.ai/env/v8-cve-2024-6100/
14.2 Code / model hubs
Section titled “14.2 Code / model hubs”https://github.com/Continuum-AI-Corp/OrcaRouter-Lite,https://github.com/continuum-ai-corp,https://github.com/abliteration-ai,https://github.com/abliterationai/abliteration-examples,https://github.com/abliteration-ai/ai-sdk-provider,https://huggingface.co/orcarouter,https://huggingface.co/abliterationaiorg,https://huggingface.co/dealignai/GLM-5.3-ABLITERATED-NVFP4,https://huggingface.co/Securelayer7/Qwen3.8-27B-Uncensored-Abliterated,https://huggingface.co/huihui-ai/BaronLLM_Offensive_Security-abliterated-GGUF,https://ollama.com/orcarouter,https://mastra.ai/models/providers/orcarouter,https://mastra.ai/models/providers/abliteration-ai,https://www.promptfoo.dev/docs/providers/orcarouter/,https://docs.privategpt.dev/providers/orcarouter- OpenRouter appendix (name-confusion only):
https://openrouter.ai/models,/pricing,/terms,https://openrouter.ai/cognitivecomputations/dolphin-mistral-24b-venice-edition:free
14.3 Press / third-party / reviews
Section titled “14.3 Press / third-party / reviews”https://techcrunch.com/2026/09/03/abliteration-ai-is-making-a-business-out-of-removing-ai-guardrails/,https://tech.yahoo.com/ai/deals/articles/abliteration-ai-making-business-removing-183757546.html,https://blog.codercops.com/blog/abliteration-ai-uncensored-models-enterprise-risk-2026,https://magica.com/news/abliteration-ai-hosted-guardrail-removed-models,https://www.explainx.ai/blog/abliteration-ai-glm-5-3-hosted-uncensored-cyber-model-2026,https://www.lesswrong.com/posts/BShGBvtxGoaZvCqBk/abliterated-models-are-now-served-cheaply-and-conveniently-1,https://www.chosun.com/english/industry-en/2026/09/10/RUI2WFD76VETHJSGFFP7QFKK6Q/https://www.reddit.com/r/opencodeCLI/comments/1vr1uuz/has_anyone_tried_orcarouter/,https://www.reddit.com/r/LocalLLaMA/comments/1f07b4b/abliteration_fails_to_uncensor_models_while_it,https://www.reddit.com/r/ArtificialIntelligence/comments/1vjtx7a/cyberkimi_just_dropped_strong_results_on_one_of/,https://www.reddit.com/r/LocalLLM/comments/1vjuhp7/cyberkimi_just_dropped_strong_results_on_one_of/,https://www.reddit.com/r/ollama/comments/1vjuksn/cyberkimi_just_dropped_strong_results_on_one_of/,https://www.reddit.com/r/LocalLLM/comments/1wrytrx/introducing_cyberpvp_cyberkimi_vs_altar1_on_100/,https://x.com/OrcaRouter,https://x.com/abliteration_ai/status/2094458081451393287,https://x.com/lordx64,https://x.com/k2sbhai/status/2089358084627955921https://radar.offseq.com/threat/ai-model-cyberkimi-claims-it-turned-a-3-day-old-v8-patch-into-a-live-chrome-exploit-in-under-24-hours-edec50aa81421de5,https://www.kobaran.com/cyberkimi-ai-claims-it-weaponized-a-chrome-v8-patch-in-under-a-day/,https://www.linkedin.com/posts/dkucinic_expanding-daybreak-as-the-cyber-defense-window-activity-7493633709191991296-QLz6,https://securitywall.co/blog/llm-security-audit-cost- Funding/acquisition:
https://www.prnewswire.com/news-releases/orcarouter-launches-the-open-llm-api-router--zero-markup-mit-licensed-100-models-302766356.html
14.4 DreadNode sources
Section titled “14.4 DreadNode sources”https://dreadnode.io/,/platform/,/pricing/,/company/about/,/company/policy/,/company/newsroom/series-a/,/company/newsroom/dreadnode-v2-security-agent-infrastructure-platform/,https://docs.dreadnode.io,/getting-started/quickstart/,/getting-started/authentication/,/platform/credits/,/ai-red-teaming/,/integrations/,/guides/building-a-capability/,https://app.dreadnode.io/?mode=register,https://app.dreadnode.io/terms,https://app.dreadnode.io/privacy,https://trust.dreadnode.io/,https://status.dreadnode.io,https://dreadnode.io/research/dreadindex/https://github.com/dreadnode,https://github.com/dreadnode/rigging,https://github.com/dreadnode/sdk,https://github.com/dreadnode/burpference,https://github.com/dreadnode/ares,https://pypi.org/project/dreadnode/,https://arxiv.org/html/2504.19855v2,https://x.com/dreadnode/status/2036164368304578640,https://www.decibel.vc/articles/dreadnode-why-great-offense-drives-defense-in-ai-security,https://www.securityweek.com/offensive-ai-startup-dreadnode-secures-14m-to-stress-test-ai-systems/,https://cybersectools.com/tools/dreadnode-crucible,https://boschko.ca/adversarial-ml/
14.5 OpenAI / Anthropic comparison sources
Section titled “14.5 OpenAI / Anthropic comparison sources”OpenAI cyber models:
https://developers.openai.com/api/docs/models/gpt-5.6-cyberhttps://www.securityweek.com/openai-unveils-new-cybersecurity-model-gpt-5-6-cyberhttps://www.csoonline.com/article/4207896/openai-launches-gpt-5-6-cyber-as-ai-narrows-vulnerability-response-window.htmlhttps://www.cnbc.com/2026/05/07/openai-rolls-out-new-gpt-5point5-cyber-to-vetted-cybersecurity-teams.htmlhttps://www.aisi.gov.uk/blog/our-evaluation-of-openais-gpt-5-5-cyber-capabilitieshttps://help-lb.openai.com/en/articles/20001259-trusted-access-for-cyber-common-issues-and-troubleshootinghttps://openai.com/index/expanding-daybreak-as-the-cyber-defense-window-narrows/https://openai.com/index/daybreak-securing-the-world
Anthropic cyber models:
https://www.anthropic.com/glasswinghttps://www.anthropic.com/claude/mythoshttps://www.anthropic.com/claude/fablehttps://www.anthropic.com/news/redeploying-fable-5https://www.anthropic.com/news/expanding-project-glasswinghttps://www.aisi.gov.uk/blog/our-evaluation-of-claude-mythos-previews-cyber-capabilitieshttps://platform.claude.com/docs/en/models/overview
14.6 Techniques research
Section titled “14.6 Techniques research”https://aisle.com/blog/ai-cybersecurity-after-mythos-the-jagged-frontierhttps://www.anthropic.com/engineering/effective-context-engineering-for-ai-agentshttps://arxiv.org/html/2402.17230v1https://arxiv.org/html/2603.22577v1https://www.mdpi.com/2076-3417/15/16/9096https://github.com/greydgl/pentestgpthttps://github.com/Yuning-J/CVE-KGRAGhttps://arxiv.org/html/2506.07468v3https://openai.com/index/unlocking-self-improvement-gpt-red/https://arxiv.org/html/2609.10316https://docs.abliteration.ai/what-is-abliterationhttps://github.com/jimbozhang/heretichttps://github.com/elder-plinius/OBLITERATUShttps://kth.diva-portal.org/smash/get/diva2:2060365/FULLTEXT01.pdfhttps://www.reddit.com/r/LocalLLaMA/comments/1u6qw5b/we_trained_a_cybersecurityfocused_mythos_like_llm/
14.7 Local reports
Section titled “14.7 Local reports”ai-security-vulnerabilities-2026-09-27.md(Part I of this manual)omo-alternatives-opencode-v2-2026-09-28.mdmulti-agentic-workflows-2026-09-28.md
Report v4: 2026-09-28 — reformatted with proper numbering for all sections and TOC. Added §12 DreadNode, §13 OpenAI/Anthropic comparison, §14 Sources. Method: multi-agent parallel deep research (5 subagents total) + TinyFish/Exa/You.com/Firecrawl + direct vendor docs fetch + ToS/privacy review. All benchmarks vendor-reported unless noted. Re-check live pricing before proposals. Authorized testing only.