“Official sources only” is a claim that needs proof. Here it is. Every academic paper, official documentation page, company technical report, and institutional publication cited across the AI Atlas — listed by guide, linked directly. 116 sources across 21 guides. All verified April 2026.
Every factual claim in the AI Atlas is grounded in one of the following: a peer-reviewed or preprint academic paper (linked to arXiv or the original journal), official company documentation or technical reports (linked to the company’s own domain), or verified institutional research from government bodies, UNESCO, or academic institutions. Opinion pieces, blog posts, and secondary sources are not used as primary citations.
Where a company has not published technical papers — Midjourney, ElevenLabs, and Runway do not — this is stated explicitly in the guide. The source listed is the most authoritative available: company documentation, API references, or research published on the company’s own domain.
The foundational peer-reviewed papers that underpin modern AI — cited in the technical depth sections of each guide. All links go directly to arXiv or the original journal.
Official documentation, API references, company technical reports, and institutional guidance — organised by the guide in which each source is cited. Click any guide name to go directly to that guide.
| Guide | Source description | Domain |
|---|---|---|
| AI Assistants | ||
| ChatGPT | openai.com/chatgpt/pricing | openai.com |
| "GPT-4 Technical Report" — OpenAI, 2023 | arxiv.org | |
| "Training Language Models to Follow Instructions" (InstructGPT) — Ouyang et al., 2022 | arxiv.org | |
| openai.com/index/hello-gpt-4o | openai.com | |
| openai.com/index/learning-to-reason-with-llms | openai.com | |
| platform.openai.com/docs/overview | platform.openai.com | |
| openai.com/safety/preparedness | openai.com | |
| openai.com/chatgpt | openai.com | |
| OpenAI Research | openai.com | |
| API Documentation | platform.openai.com | |
| Claude | anthropic.com/claude | anthropic.com |
| docs.anthropic.com | docs.anthropic.com | |
| "Constitutional AI: Harmlessness from AI Feedback" — Bai et al., 2022 | arxiv.org | |
| transformer-circuits.pub | transformer-circuits.pub | |
| transformer-circuits.pub | transformer-circuits.pub | |
| anthropic.com/responsible-scaling-policy | anthropic.com | |
| anthropic.com/claude/model-cards | anthropic.com | |
| anthropic.com/news/computer-use | anthropic.com | |
| docs.anthropic.com/en/docs/about-claude/models | docs.anthropic.com | |
| Interpretability research | transformer-circuits.pub | |
| Gemini | gemini.google.com | gemini.google.com |
| one.google.com | one.google.com | |
| "Mixtral of Experts" — Jiang et al., 2023 | arxiv.org | |
| "Gemma: Open Models Based on Gemini Research" — Google DeepMind, 2024 | arxiv.org | |
| blog.google/technology/google-deepmind | blog.google | |
| ai.google.dev/gemini-api/docs | ai.google.dev | |
| Google AI Developer | ai.google.dev | |
| Microsoft Copilot | microsoft.com/copilot | microsoft.com |
| learn.microsoft.com — Copilot Architecture | learn.microsoft.com | |
| learn.microsoft.com — Copilot Privacy | learn.microsoft.com | |
| learn.microsoft.com/azure/ai-services/openai | learn.microsoft.com | |
| copilot.microsoft.com | copilot.microsoft.com | |
| M365 Copilot docs | learn.microsoft.com | |
| Perplexity | perplexity.ai/pro | perplexity.ai |
| perplexity.ai/hub/blog | perplexity.ai | |
| docs.perplexity.ai | docs.perplexity.ai | |
| perplexity.ai | perplexity.ai | |
| Perplexity blog | perplexity.ai | |
| Grok | github.com/xai-org/grok-1 | github.com |
| x.ai/blog/grok-os | x.ai | |
| x.ai/blog/grok-2 | x.ai | |
| docs.x.ai | docs.x.ai | |
| grok.com | grok.com | |
| xAI blog | x.ai | |
| Open Source Models | ||
| Llama | "LLaMA: Open and Efficient Foundation Language Models" — Touvron et al., 2023 | arxiv.org |
| "Llama 2: Open Foundation and Fine-Tuned Chat Models" — Touvron et al., 2023 | arxiv.org | |
| "Gemma 2: Improving Open Language Models" — Google DeepMind, 2024 | arxiv.org | |
| github.com/meta-llama/llama-models | github.com | |
| ai.meta.com/blog/llama-4 | ai.meta.com | |
| "Evaluating Large Language Models Trained on Code" (Codex) — Chen et al., 2021 | arxiv.org | |
| "Direct Preference Optimization" — Rafailov et al., 2023 | arxiv.org | |
| github.com/meta-llama/llama-models — LICENSE | github.com | |
| ai.meta.com/llama | ai.meta.com | |
| GitHub — model cards | github.com | |
| Ollama (run locally) | ollama.com | |
| Hugging Face PEFT | huggingface.co | |
| Mistral | docs.mistral.ai | docs.mistral.ai |
| "Mistral 7B" — Jiang et al., 2023 | arxiv.org | |
| "Mixtral 8x7B — Sparse Mixture of Experts" — Jiang et al., 2024 | arxiv.org | |
| mistral.ai | mistral.ai | |
| Le Chat | chat.mistral.ai | |
| Image, Video and Audio AI | ||
| Midjourney | midjourney.com/account | midjourney.com |
| "High-Resolution Image Synthesis with Latent Diffusion Models" — Rombach et al., 2022 | arxiv.org | |
| "Learning Transferable Visual Models From Natural Language Supervision" (CLIP) — Radford et al., 202 | arxiv.org | |
| docs.midjourney.com | docs.midjourney.com | |
| midjourney.com | midjourney.com | |
| Community showcase | midjourney.com | |
| DALL-E 3 | openai.com/dall-e-3 | openai.com |
| openai.com/papers/dall-e-3.pdf | cdn.openai.com | |
| platform.openai.com/docs/guides/images | platform.openai.com | |
| Sora | sora.com | sora.com |
| openai.com/research/video-generation-models-as-world-simulators | openai.com | |
| ElevenLabs | elevenlabs.io/pricing | elevenlabs.io |
| elevenlabs.io/docs | elevenlabs.io | |
| elevenlabs.io/docs/models | elevenlabs.io | |
| elevenlabs.io | elevenlabs.io | |
| Safety policy | elevenlabs.io | |
| Runway | runwayml.com/pricing | runwayml.com |
| runwayml.com/research/gen-3-alpha | runwayml.com | |
| docs.runwayml.com | docs.runwayml.com | |
| runwayml.com | runwayml.com | |
| Research blog | runwayml.com | |
| Productivity AI | ||
| Copilot in M365 | microsoft.com/microsoft-365/copilot | microsoft.com |
| learn.microsoft.com — Copilot Architecture | learn.microsoft.com | |
| data protection commitments | learn.microsoft.com | |
| learn.microsoft.com — Copilot Studio | learn.microsoft.com | |
| Microsoft Learn docs | learn.microsoft.com | |
| Gemini for Workspace | workspace.google.com/products/gemini | workspace.google.com |
| workspace.google.com/security | workspace.google.com | |
| notebooklm.google.com/about | notebooklm.google.com | |
| NotebookLM | notebooklm.google.com | |
| Vertex AI | cloud.google.com | |
| Notion AI | notion.so/pricing | notion.so |
| notion.so/blog/notion-ai | notion.so | |
| notion.so/help/notion-ai | notion.so | |
| notion.so privacy overview | notion.so | |
| notion.so | notion.so | |
| Understanding AI | ||
| AI History | Mind, Vol. LIX, No. 236 | academic.oup.com |
| "Attention Is All You Need" — Vaswani et al., 2017 | arxiv.org | |
| "Language Models are Few-Shot Learners" (GPT-3) — Brown et al., 2020 | arxiv.org | |
| nature.com | nature.com | |
| NeurIPS proceedings | proceedings.neurips.cc | |
| "BERT: Pre-training of Deep Bidirectional Transformers" — Devlin et al., 2018 | arxiv.org | |
| "Training Language Models to Follow Instructions" (InstructGPT) — Ouyang et al., 2022 | arxiv.org | |
| Start Here | "Attention Is All You Need" — Vaswani et al., 2017 | arxiv.org |
| "Training Language Models to Follow Instructions" (InstructGPT) — Ouyang et al., 2022 | arxiv.org | |
| "Language Models are Few-Shot Learners" (GPT-3) — Brown et al., 2020 | arxiv.org | |
| "Scaling Laws for Neural Language Models" — Kaplan et al., 2020 | arxiv.org | |
| "Constitutional AI: Harmlessness from AI Feedback" — Bai et al., 2022 | arxiv.org | |
| AI Glossary | "Constitutional AI: Harmlessness from AI Feedback" — Bai et al., 2022 | arxiv.org |
| "Attention Is All You Need" — Vaswani et al., 2017 | arxiv.org | |
| "Training Compute-Optimal Large Language Models" (Chinchilla) — Hoffmann et al., 2022 | arxiv.org | |
| "Learning Transferable Visual Models From Natural Language Supervision" (CLIP) — Radford et al., 202 | arxiv.org | |
| Use Cases and For You | ||
| AI for Students | qaa.ac.uk — Academic Integrity | qaa.ac.uk |
| AI for Teachers | gov.uk | gov.uk |
| UNESCO | unesdoc.unesco.org | |
Sources are reviewed and updated as guides are revised. All links on this page were verified in April 2026. If you find a broken link, a missing citation, or believe any claim is incorrectly attributed, please contact us via the About page.