Company profiles
The AI companies on the record
Scored 0 to 100 on the gains and problems documented about each one. A high score means a lot is documented, in both directions.
- 10
- companies tracked
- 95
- top score, OpenAI
- 2026-11-10
- data as of
Ranking of tracked AI companies
- 1

OpenAI
Maker of ChatGPT, Codex and the GPT model line (GPT-6 launched in fall 2026), with a large consumer and developer surface. The Sora video app was shut down in April 2026.
95Very high impact score296137 gains · 159 problems75272 independent sources · 0 backed by strong evidenceHigh confidence Falling attention
- 2

Google DeepMind
Google's consolidated AI research and product arm: Gemini (Gemini 4 launched October 2026), Gemini Notebook (formerly NotebookLM), AI Overviews in Search, and AlphaFold.
80Very high impact score14466 gains · 78 problems67140 independent sources · 0 backed by strong evidenceHigh confidence Falling attention
- 3
Anthropic
Maker of Claude, Claude Code and Claude Cowork, with growing enterprise and developer-tool distribution.
69High impact score8736 gains · 51 problems63104 independent sources · 0 backed by strong evidenceHigh confidence Falling attention
- 4

Microsoft AI
Microsoft Copilot, GitHub Copilot, Microsoft's in-house MAI models, and the Foundry cloud platform distribute AI across workplace, developer, health, and cloud products.
49Middling impact score3714 gains · 23 problems4834 independent sources · 1 backed by strong evidenceHigh confidence Stable attention
- 5
Meta AI
Meta's assistant and models, now led by the Muse family (Muse Spark), alongside Llama open models and AI features across Facebook, Instagram, WhatsApp and Ray-Ban Meta glasses.
45Middling impact score2610 gains · 16 problems4725 independent sources · 0 backed by strong evidenceHigh confidence Stable attention
- 6
xAI (SpaceXSI)
SpaceX's AI division: founded as xAI, renamed SpaceXAI in July 2026 and branded SpaceXSI in October 2026. Maker of Grok, owner of X, and owner of the Cursor coding agent since SpaceX closed its $60 billion acquisition of Anysphere in August 2026.
42Middling impact score219 gains · 12 problems4623 independent sources · 0 backed by strong evidenceHigh confidence Stable attention
- 7
Perplexity
An AI-native answer engine and the Comet browser, operating in the search and publishing ecosystem.
33Low impact score154 gains · 11 problems3914 independent sources · 0 backed by strong evidenceHigh confidence Stable attention
- 8
Nvidia
The compute layer behind much of frontier-model training and deployment: Blackwell and Vera Rubin systems, CUDA, and the Nemotron, Cosmos and GR00T model families.
20Low impact score81 gains · 7 problems3210 independent sources · 0 backed by strong evidenceHigh confidence Rising attention
- 9
Mistral AI
A European model lab with open-weight releases, enterprise deployments, and the Mistral Vibe assistant (formerly Le Chat).
20Low impact score64 gains · 2 problems306 independent sources · 0 backed by strong evidenceHigh confidence Falling attention
- 10

Cohere
An enterprise-focused model and retrieval provider (Command models, the North agent platform) with limited consumer visibility.
4Low impact score10 gains · 1 problems214 independent sources · 0 backed by strong evidenceModerate confidence Rising attention
Before you read too much into thisThese scores only count the evidence Truvace has on record, run through the formula on our methodology page. They don't capture everything a company does, and they aren't investment advice. A higher problem score means more problems have been documented. That isn’t the same as a company being less safe.
ace