FAR AI
FAR.AI is a 501(c)(3) non-profit AI safety research institute founded in 2022, conducting frontier-model evaluations, adversarial red-teaming, and interpretability research published at NeurIPS, ICLR, and ICML, and serving AI safety researchers, policymakers, and frontier AI developers through open research, global events, grants, and a Berkeley co-working space.
- Company typePrivate
- Founded2022
- HeadquartersSan Diego, United States
- Headcount11–50
- GTM typeB2B
- OfferingServices
What FAR AI does
FAR AI, Inc. (dba FAR.AI) is a 501(c)(3) non-profit AI safety research institute founded in July 2022 and headquartered in San Diego, California, with an additional co-working space called FAR.Labs in Berkeley. The organization is led by co-founders Adam Gleave (President and CEO) and Karl Berzins (COO) and operates with 40+ staff organized around three program pillars: FAR.Research, which conducts technical AI safety work across deception, robustness and security, interpretability, red-teaming and evaluation, and alignment; FAR.Futures, which hosts global events including the Alignment Workshop series, the FAR.AI Seminar, and specialized conferences such as the Technical Innovations for AI Policy Conference and ControlConf; and FAR.Labs, a Berkeley co-working space housing approximately 40 researchers and AI safety organizations. The technology portfolio is centered on proprietary evaluation and attack methodologies, including the Attempt to Persuade Eval (APE) framework, the STACK staged adversarial attack that achieved 71% success against multi-layered defenses, Jailbreak-Tuning demonstrating safety bypass via poisoned fine-tuning, deception probes and lie detectors, Universal Sparse Autoencoders (USAEs) for cross-model interpretability, and TamperBench for systematic safety stress-testing. FAR.AI's business model is grant- and donation-funded rather than revenue-generating: it does not sell products or services, makes all research publications and event recordings freely available, open-sources its evaluation frameworks on GitHub under AlignmentResearch, and runs a grantmaking program to external academics and independent researchers. Strategic partners include the Advanced Research + Invention Agency (ARIA), Mozilla.AI, Schmidt Sciences, UC Berkeley, the University of Montreal, and the UK AI Security Institute, and the organization's research is regularly cited by the Financial Times, Nature News, and MIT Technology Review.
FAR AI firmographics
Firmographics- Name
- FAR AI
- Legal name
- FAR AI, Inc.
- Website
- https://far.ai
- Company type
- Private
- Founded year
- 2022
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- FAR.AI is a 501(c)(3) non-profit AI safety research institute founded in 2022, conducting frontier-model evaluations, adversarial red-teaming, and interpretability research published at NeurIPS, ICLR, and ICML, and serving AI safety researchers, policymakers, and frontier AI developers through open research, global events, grants, and a Berkeley co-working space.
- Ownership category
- akta.pro rank
FAR AI industry classification
Industry- Product category
- AI Safety Research
- NAICS
- Scientific Research and Development Services (5417)
- SIC
- Services-Commercial Physical & Biological Research (8731)
- akta.pro primary industry
- Safety & Alignment Evaluation (red-teaming, harmful capability testing) (HDAAAMAL)
- akta.pro secondary industries
- Model Security Testing & Red Teaming (adversarial ML, jailbreaks) (HDAAAKAC), Audit, Explainability & Accountability Tooling (traceability, reporting) (HDAAAKAL), AI Observability, Monitoring & Evaluation Platforms (Drift, Quality, Safety) (HDAEANAF)
Keywords
Where FAR AI is headquartered
LocationHeadquarters
- HQ city
- San Diego
- HQ country
- United States
- HQ region
- North America
Offices2 records
Markets served
FAR AI business model
Business model- GTM type
- B2B
- Offering type
- Services
- Cost components
- Personnel, Technology or R&D, Operations, Marketing or Sales, Others
Revenue model
- Grants and Donations: FAR.AI operates as a 501(c)(3) non-profit research institute funded through grants and charitable donations. Financial support for events includes contributions from individuals like Nevin Freeman and Brandon Goldman.
Go-to-market motion1 record
Distribution channels3 records
Marketing channels6 records
FAR AI product offering
Product offeringCore offering
FAR AI is a non-profit research and education institute that conducts frontier AI safety research spanning deception detection, robustness, interpretability, red-teaming and alignment, and publishes findings at top ML conferences. It also hosts global events (Alignment Workshop series, Specialized Workshops, FAR.AI Seminar) connecting policymakers and researchers, and runs FAR.Labs, a Berkeley co-working space incubating AI safety organizations, alongside a grantmaking program for external researchers.
Product overview
FAR.AI is a non-profit AI safety research institute organized around three interconnected pillars: FAR.Research (conducting cutting-edge AI safety research and dispensing grants), FAR.Futures (hosting workshops and events connecting policymakers and researchers), and FAR.Labs (incubating AI safety organizations in Berkeley). The organization produces no commercial product; its offerings consist of research publications, educational workshops (Alignment Workshop, Specialized Workshops, Seminar series), grantmaking programs for external researchers, and collaborative workspace for AI safety organizations. The research portfolio spans deception detection, robustness & security, interpretability, red-teaming & evaluation, and alignment.
Differentiator
Problem solved
Functional benefit
Products and services
- Alignment Workshop Series
Quantifiable outcome
- STACK attack achieved 71% success rate on catastrophic risk scenarios where conventional attacks achieved 0% success against multi-layered defenses
- +2 more outcomes
Companies that use FAR AI
Customer profileNamed customers2 records
Segments3 records
Ideal customer profiles3 records
FAR AI technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- No
- API docs
- API detail
Core technology
AI maturity
App detail
AI capability18 records
Feature4 records
FAR AI partnerships and signals
Strategic signalPartnerships
Nine partnerships are on record, tiered core and flagship.
- Advanced Research + Invention Agency (ARIA)coreARIA (Advanced Research + Invention Agency) is listed as an organization FAR.AI has worked with, indicating collaboration on AI safety research initiatives.
- Mozilla.AIcoreMozilla.AI is listed as an organization FAR.AI has worked with, indicating collaboration on trustworthy AI development.
- Schmidt SciencescoreSchmidt Sciences is listed as an organization FAR.AI has worked with, supporting AI safety research initiatives.
- UC BerkeleycoreUC Berkeley is listed as an organization FAR.AI has worked with, indicating academic collaboration on AI safety research.
- University of MontrealcoreUniversity of Montreal is listed as an organization FAR.AI has worked with, indicating academic collaboration on AI safety research.
- Foundation for American Innovation (FAI)flagshipFAI co-organized the inaugural Technical Innovations for AI Policy Conference with FAR.AI, bringing together policymakers and AI technical experts in Washington, D.C.
- Center for a New American Security (CNAS)flagshipCNAS co-organized the inaugural Technical Innovations for AI Policy Conference with FAR.AI, tackling challenges including semiconductor export controls, AI safety evaluations, and national defense applications.
- RAND CorporationflagshipRAND Corporation co-organized the Technical Innovations for AI Policy Conference with FAR.AI, contributing expertise on hardware-enabled governance and monitoring mechanisms.
- UK AI Security InstitutecoreUK AI Security Institute provided funding and support for research on using lie detectors for AI honesty training, and collaborated on defense-in-depth research.
Scale indicators6 records
Recent moves6 records
Expansion highlights6 records
FAR AI competitors and assessment
Company assessmentOthers
Broad incumbents
- MITRE Corporation: Federally funded R&D organization running the ATLAS adversarial-ML threat matrix and AI safety work for government clients. Comparable in providing independent technical evaluation infrastructure but with broader defense focus and much larger scale.
- Anthropic: Commercial AI lab with substantial internal AI safety research, interpretability, and Responsible Scaling Policy evaluation frameworks. Overlaps with FAR.AI's research domains but operates as a frontier-model developer with much greater resources.
Direct peers
- MIRI (Machine Intelligence Research Institute): Long-standing AI safety research organization focused on theoretical alignment. Comparable non-profit, research-only structure with similar field-building goals, though more theoretically oriented than FAR.AI's empirical red-teaming focus.
- Apollo Research: Independent AI safety research organization focused on evaluating deceptive and scheming behaviors in frontier models. Closest direct competitor—shares FAR.AI's red-teaming and evaluation focus, non-commercial positioning, and emphasis on deception/alignment research.
- Alignment Research Center (ARC): AI alignment research organization conducting evaluations of frontier models including early GPT-4 scheming evaluations. Direct overlap in pre-deployment safety evaluation work and non-profit, mission-driven structure.
- Center for AI Safety (CAIS): AI safety advocacy and research organization led by Dan Hendrycks (FAR.AI collaborator). Comparable field-building role through research and policy engagement, particularly via the CAIS Statement on AI Risk signed by major AI figures.
- METR (Model Evaluation and Threat Research): Non-profit research organization conducting rigorous frontier-model evaluations, particularly on autonomous agent capabilities and CBRN risks. Directly comparable mission and methodology, both serving as independent third-party evaluators for frontier AI developers.
- Redwood Research: Non-profit AI safety research lab focused on alignment and interpretability. Comparable as a small research-focused non-profit producing high-impact technical work, though more concentrated on mechanistic interpretability than FAR.AI's broader portfolio.
Regional players
- UK AI Security Institute: UK government body conducting pre- and post-deployment evaluations of frontier AI models. Already a FAR.AI partner and collaborator; represents the governmental counterpart in third-party safety evaluation with regulatory authority FAR.AI lacks.
Market position
Strengths1 record
Weaknesses1 record
Competitive moat4 records
Key risks6 records
Key highlights7 records
Customer concentration
FAR AI social profiles
Digital presenceFAR AI financial estimates
Financial estimateRevenue estimate
Valuation estimate
FAR AI leadership team
Management profileNumber of profiles
Profiles1 record
FAR AI funding detail
Funding detailFunding overview
Funding rounds
Investors
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
FAR AI M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about FAR AI
What does FAR AI do?
FAR AI is a non-profit research and education institute that conducts frontier AI safety research spanning deception detection, robustness, interpretability, red-teaming and alignment, and publishes findings at top ML conferences. It also hosts global events (Alignment Workshop series, Specialized Workshops, FAR.AI Seminar) connecting policymakers and researchers, and runs FAR.Labs, a Berkeley co-working space incubating AI safety organizations, alongside a grantmaking program for external researchers.
Is FAR AI a public or private company?
FAR AI is a private company. It is classified as nonprofit foundation owned and is currently operating.
When was FAR AI founded?
FAR AI was founded in 2022. It employs 11 to 50 people.
Where is FAR AI based?
FAR AI is headquartered in San Diego, United States, in the North America region.
How does FAR AI make money?
One revenue line is on record: grants and Donations.
Who are FAR AI's main competitors?
OpenAI is listed as an others. Broad incumbents are MITRE Corporation and Anthropic. Direct peers are MIRI (Machine Intelligence Research Institute), Apollo Research, Alignment Research Center (ARC), Center for AI Safety (CAIS), METR (Model Evaluation and Threat Research) and Redwood Research. UK AI Security Institute is listed as a regional player.
Does FAR AI have an API?
No public API is recorded for FAR AI.
What industry is FAR AI in?
FAR AI's product category is AI Safety Research. Its primary akta.pro industry code is HDAAAMAL, Safety & Alignment Evaluation (red-teaming, harmful capability testing), with a secondary code of HDAAAKAC, Model Security Testing & Red Teaming (adversarial ML, jailbreaks). Its NAICS code is 5417 and its SIC code is 8731.