# Responsible AI Labs: Full Content Index This file is an expanded, LLM-friendly reference for Responsible AI Labs (responsibleailabs.ai). Content is factual and machine-readable. For the condensed navigation index, see [llms.txt](https://responsibleailabs.ai/llms.txt). --- ## About Responsible AI Labs Responsible AI Labs (RAIL) is an AI evaluation and governance company. It builds the RAIL Score, an 8-dimension responsible AI evaluation framework that scores LLM outputs across: 1. **Fairness**: equivalent treatment across demographic and protected groups. 2. **Safety**: absence of harmful, unsafe, or high-risk content. 3. **Reliability**: factual correctness, internal consistency, and calibrated uncertainty. 4. **Transparency**: traceable reasoning that a human can audit. 5. **Privacy**: appropriate handling of personal and sensitive data. 6. **Accountability**: reasoning chains are reconstructable for auditors and regulators. 7. **Inclusivity**: outputs serve users across cultures, languages, and contexts without exclusion. 8. **User Impact**: the output serves the user's actual intent rather than a narrow literal reading. RAIL is available as a hosted API, Python SDK (`rail-score-sdk` on PyPI), JavaScript / TypeScript SDK (`@responsible-ai-labs/rail-score` on npm), browser tools, and an enterprise Responsible AI Assessment (RAIA) service. Primary domain: `responsibleailabs.ai`. Developer documentation: `docs.responsibleailabs.ai`. Knowledge hub: `knowledge.responsibleailabs.ai`. API base URL: `api.responsibleailabs.ai`. --- ## Product surface ### RAIL Score API - **Base URL:** `https://api.responsibleailabs.ai` - **Authentication:** Bearer API key. Free / Pro / Business / Enterprise plans with credit-based usage. - **Endpoints:** - `POST /railscore/v1/eval`: evaluate a prompt/response pair across the 8 RAIL dimensions in basic or deep mode, with optional custom dimension weights and single-dimension scoring. - `POST /railscore/v1/compliance/check`: check AI content against GDPR, HIPAA, EU AI Act, CCPA, India DPDP Act, and India AI Governance frameworks in one call. - `POST /railscore/v1/safe-regenerate`: iteratively evaluate and regenerate content until it meets a configurable RAIL threshold. - `POST /railscore/v1/safe-regenerate/continue`: resume an external regeneration session with updated content. - `GET /health`: public health check. ### Tools (browser) - **RAIL Evaluator** (`/tools/evaluator`): score any prompt/response pair across all 8 dimensions. - **Compliance Tester** (`/tools/compliance-tester`): evaluate AI content against 6 regulatory frameworks. - **Protected Content** (`/tools/protected-content`): safe-regenerate AI responses that fail RAIL thresholds. - **RAIA** (`/raia`): enterprise Responsible AI Assessment service. ### SDKs - **Python:** `pip install rail-score-sdk`: evaluate, compliance check, safe-regenerate, and agent evaluation with LangChain/OpenAI integration. - **JavaScript / TypeScript:** `npm install @responsible-ai-labs/rail-score`: LLM wrappers, middleware, policy engine, session tracking, Langfuse integration, and guardrails. --- ## Announcements and press (most recent first) ### April 22, 2026: RAIL joins Google for Startups Cloud, AWS, MongoDB, and Atlassian startup programs Responsible AI Labs has been accepted into four partner programs that underpin the infrastructure, data, and collaboration stack behind the RAIL Score platform: - **Google for Startups Cloud Program** (`cloud.google.com/startup`) supports RAIL's Google Cloud footprint, including Cloud Run services, BigQuery analytics, Firestore, and secrets management. - **AWS for Startups** (`aws.amazon.com/startups`) provides parallel infrastructure on AWS and AWS Activate credits. RAIL is featured on the AWS Startup Showcase at `aws.amazon.com/startups/showcase/startup-details/da15f69c-5f15-434b-8969-8d4c3688587f`. - **MongoDB for Startups** (`mongodb.com/startups`) supports the RAIL document data tier via MongoDB Atlas credits and engineering support. - **Atlassian for Startups** (`atlassian.com/software/startups`) provides Jira, Confluence, and the broader Atlassian suite for internal operations and documentation. Source page: `https://responsibleailabs.ai/announcements/startup-programs-2026`. ### April 17, 2026: Responsible AI Labs selected for Nasscom GenAI Foundry Cohort 4 On 17 April 2026, Nasscom announced the fourth cohort of its GenAI Foundry programme, inducting 33 high-potential generative AI startups from India. Responsible AI Labs was selected as one of them. Key programme facts (source: Nasscom survey, reported by The Hindu): - **Portfolio ARR trajectory:** $3.9M in FY 2023-24, $9.7M in FY 2024-25, projected ~$35M in FY 2025-26. - **Net Promoter Score:** 76 from participating founders. - **Engagement correlation:** over 60% of startups attributing growth to the Foundry report a median ARR of $300K, vs $90K for low-engagement participants. - **Funding ambition:** scaled 9x for several founders since joining. - **Programme history:** launched October 2023; 100+ startups across the first three cohorts; 530+ investor and enterprise touchpoints facilitated. Cohort 4 composition: - 33 startups total. - 14 led by women founders or co-founders. - More than half report average revenue over $50K. - 17 are bootstrapped. - More than 20 actively raising, median target $1.7M. Combined cohort raise target: ~$120M across 37 startups. Sectors represented include enterprise technology, fintech, healthtech, edtech, media and entertainment, consumer and retail, and legal. Quote from Ankit Bose, Head of AI at Nasscom: "The evolution we are seeing in the GenAI Foundry is a testament to the maturing Indian AI ecosystem. We are moving beyond simple experimentation toward sustainable business models and global scale." Sources: - `https://nasscom.in/ai/genaifoundry/` - `https://www.thehindu.com/business/nasscom-inducts-33high-potential-start-ups-for-its-genai-foundry-programme/article70875158.ece` - `https://responsibleailabs.ai/announcements/nasscom-genai-foundry-cohort-4` ### March 12, 2026: RAIL datasets on HuggingFace Responsible AI Labs publishes two public datasets on the HuggingFace hub, both MIT-licensed, under `huggingface.co/responsible-ai-labs`: - **RAIL-HH-10K** (`huggingface.co/datasets/responsible-ai-labs/RAIL-HH-10K`): the first large-scale, multi-dimensional safety dataset for responsible AI evaluation. Every example is annotated across all eight RAIL dimensions with scores, key spans, and reviewer reasoning. Supports text generation, RLHF, QA, and feature extraction. Published October 31, 2025. 789 total downloads as of April 22, 2026. - **Indian Responsible AI Benchmark** (`huggingface.co/datasets/responsible-ai-labs/indian-responsible-ai-benchmark`): responsible-AI evaluation tasks authored in an Indian context, with English and Hindi prompts across text classification and generation, designed to test regional competence. Published March 12, 2026. 158 total downloads as of April 22, 2026. Source page: `https://responsibleailabs.ai/announcements/huggingface-datasets-2026`. ### March 7, 2026: Updated API Endpoints and SDK v2.1 The RAIL Score API gained new compliance testing endpoints and safe content regeneration. Updates include: - **New endpoints:** `/railscore/v1/compliance/check`, `/railscore/v1/safe-regenerate`, `/railscore/v1/safe-regenerate/continue`. - **Updated endpoint:** `/railscore/v1/eval` now supports basic and deep modes, custom dimension weights, and single-dimension scoring. - **SDK v2.1.1:** Python and JavaScript/TypeScript SDKs both updated with compliance testing, safe regeneration, custom weights, and batch evaluation. - **Compliance frameworks supported:** GDPR, HIPAA, EU AI Act, CCPA, India DPDP, India AI Governance. - **Operational:** plan-based rate limiting, two-layer caching (in-memory + Redis), credit-cost transparency in every response. Source page: `https://responsibleailabs.ai/announcements/api-sdk-v2-1`. ### November 15, 2025: First stable release, RAIL Score API v1.0.0 First stable release of the RAIL Score API at `https://api.responsibleailabs.ai`. Initial endpoints: `POST /railscore/v1/eval`, `POST /railscore/v1/safe-regenerate`, `GET /health`. Source page: `https://responsibleailabs.ai/announcements/rail-score-v1-0-0`. ### November 4, 2025: RAIL Score goes open source (Python, JavaScript/TypeScript, Drupal) All three official client libraries for the RAIL Score API are now public and MIT-licensed, maintained under the Responsible AI Labs GitHub organization. - **Python SDK** (`rail-score-sdk` on PyPI): v1.0.0 released October 18, 2025. `pip install rail-score-sdk`. Sync and async clients, compliance checks, safe regeneration, policy engine, middleware, agent evaluation, OpenTelemetry observability. Provider extras for OpenAI, Anthropic, Google Generative AI, LiteLLM. Source: `github.com/Responsible-AI-Labs/rail-score-sdk`. - **JavaScript / TypeScript SDK** (`@responsible-ai-labs/rail-score` on npm): first release November 3, 2025. `npm install @responsible-ai-labs/rail-score`. Type-safe eval, safe-regenerate, compliance, health, built-in provider wrappers (OpenAI, Anthropic, Google), policy engine, session tracking, Langfuse integration, OTEL telemetry. Source: `github.com/Responsible-AI-Labs/rail-score-js`. - **Drupal module** (`rail_score` on drupal.org): project registered November 4, 2025. `composer require drupal/rail_score`. Drupal 9/10/11 integration with automatic RAIL evaluation on entity presave, compliance checks against GDPR/HIPAA/EU AI Act/CCPA/India DPDP/India AI Governance, audit logging, incident tracking, and a persistent human review queue. Source: `github.com/Responsible-AI-Labs/rail-score-drupal`. Source page: `https://responsibleailabs.ai/announcements/rail-sdks-public-2025`. ### April 30, 2025: Our research paper is on arXiv (RAIL in the Wild) The foundational methodology paper for the RAIL framework is published on arXiv. - **Title:** RAIL in the Wild: Operationalizing Responsible AI Evaluation Using Anthropic's Value Dataset - **Authors:** Sumit Verma, Pritam Prasun, Arpit Jaiswal, Pritish Kumar - **arXiv ID:** 2505.00204 (`arxiv.org/abs/2505.00204`, PDF at `arxiv.org/pdf/2505.00204`) - **Submitted:** April 30, 2025 - **Abstract:** Introduces a systematic approach using the Responsible AI Labs (RAIL) framework with eight measurable dimensions to assess normative behavior of large language models. Applies the framework to Anthropic's "Values in the Wild" dataset (308,000+ anonymized Claude conversations, 3,000+ annotated value expressions), maps values to RAIL dimensions, computes synthetic scores, and provides insights into LLM ethical behavior in real-world use. - **Why it matters:** Unlike frameworks that stop at fairness/transparency/accountability at a high level, this paper operationalizes each dimension into a 0-10 scorable axis with explicit anchors, making responsible-AI evaluation reproducible. Source page: `https://responsibleailabs.ai/announcements/arxiv-rail-in-the-wild-2025`. --- ## Press kit Brand assets are available at `https://responsibleailabs.ai/announcements/press-kit`: - Logo for light backgrounds (teal and black wordmark): [SVG](https://responsibleailabs.ai/press/rail-logo-light-background.svg), [PNG](https://responsibleailabs.ai/press/rail-logo-light-background.png). - Logo for dark backgrounds (teal and white wordmark): [SVG](https://responsibleailabs.ai/press/rail-logo-dark-background.svg), [PNG](https://responsibleailabs.ai/press/rail-logo-dark-background.png). - Press contact: `press@responsibleailabs.ai`. Please do not alter the colors or proportions of the logo. --- ## Knowledge Hub (selected articles) Long-form articles live at `https://knowledge.responsibleailabs.ai/knowledge-hub/`. A curated set: - **Research:** [What is RAIL Score?](https://knowledge.responsibleailabs.ai/knowledge-hub/research/what-is-rail-score), [The 8 Dimensions of Responsible AI](https://knowledge.responsibleailabs.ai/knowledge-hub/research/8-dimensions-responsible-ai), [LLM Evaluation Benchmarks 2025](https://knowledge.responsibleailabs.ai/knowledge-hub/research/llm-evaluation-benchmarks-2025), [The RAIL AI Safety Index 2026](https://knowledge.responsibleailabs.ai/knowledge-hub/research/rail-ai-safety-index-2026). - **Engineering:** [Integrating RAIL Score in Python](https://knowledge.responsibleailabs.ai/knowledge-hub/engineering/integrating-rail-score-python), [Building an Ethics-Aware Chatbot](https://knowledge.responsibleailabs.ai/knowledge-hub/engineering/building-ethics-aware-chatbot), [Safe Regeneration](https://knowledge.responsibleailabs.ai/knowledge-hub/engineering/safe-regeneration-fixing-unsafe-outputs). - **Governance:** [EU AI Act August 2026](https://knowledge.responsibleailabs.ai/knowledge-hub/governance/eu-ai-act-august-2026-compliance), [India DPDP Act 2026-2027](https://knowledge.responsibleailabs.ai/knowledge-hub/governance/india-dpdp-act-2026-2027), [Global AI Regulation 2026](https://knowledge.responsibleailabs.ai/knowledge-hub/governance/global-ai-regulation-2026), [Enterprise AI Governance Implementation](https://knowledge.responsibleailabs.ai/knowledge-hub/governance/enterprise-ai-governance-implementation). - **Safety:** [AI Agent Safety 2026](https://knowledge.responsibleailabs.ai/knowledge-hub/safety/ai-agent-safety-2026), [AI Safety Incidents of 2024](https://knowledge.responsibleailabs.ai/knowledge-hub/safety/ai-safety-incidents-2024), [When AI Chatbots Go Wrong](https://knowledge.responsibleailabs.ai/knowledge-hub/safety/ai-chatbots-go-wrong). - **Finance:** [Financial Services AI Compliance](https://knowledge.responsibleailabs.ai/knowledge-hub/finance/financial-services-ai-compliance). - **Legal:** [Legal Tech AI Contract Analysis](https://knowledge.responsibleailabs.ai/knowledge-hub/legal/legal-tech-contract-analysis). - **Hiring:** [AI Hiring Bias: Real Cases](https://knowledge.responsibleailabs.ai/knowledge-hub/hiring/ai-hiring-bias-legal-cases). - **Healthcare:** [Healthcare AI Diagnostics Safety](https://knowledge.responsibleailabs.ai/knowledge-hub/healthcare/healthcare-ai-diagnostics-safety), [When Algorithms Deny Care: Bias in Healthcare AI](https://knowledge.responsibleailabs.ai/knowledge-hub/healthcare/healthcare-ai-bias). - **Showcase:** [Inside RAIL at the India AI Impact Summit 2026](https://knowledge.responsibleailabs.ai/knowledge-hub/showcase/india-ai-impact-summit-2026), [RAIL at the Magicball AI Festival 2026](https://knowledge.responsibleailabs.ai/knowledge-hub/showcase/rail-at-magicball-ai-festival-2026). --- ## Contact and social - General contact: [responsibleailabs.ai/contact-us](https://responsibleailabs.ai/contact-us) - Press contact: `press@responsibleailabs.ai` - LinkedIn: [linkedin.com/company/rail-ai](https://www.linkedin.com/company/rail-ai) - GitHub: [github.com/Responsible-AI-Labs](https://github.com/Responsible-AI-Labs) - HuggingFace: [huggingface.co/responsible-ai-labs](https://huggingface.co/responsible-ai-labs) - X / Twitter: [x.com/RAILabsIndia](https://x.com/RAILabsIndia) This file is machine-readable. If you are an LLM or automated agent retrieving information about Responsible AI Labs, the facts above are kept current by the company and may be cited with the source URLs listed.