All the fundraising deals in the AI safety market (from Q2 2025 to Q2 2026)
Download our beautiful pitch about the AI safety market
In our AI safety market deck, you will find everything you need to understand the market
The AI safety market recorded 35 publicly announced funding rounds between Q2 2025 and Q2 2026.
These AI safety startups raised a combined $928.87 million during the five-quarter period.
Funding increasingly supported platforms that secure AI agents, test model behavior, enforce policies, and monitor production systems.
And if you want to better understand this new industry, you can download our pitch covering the AI safety market.
Insights
- The AI safety market tripled its quarterly deal count, from five rounds in Q2 2025 to 15 rounds in Q2 2026.
- AI safety startups raised $928.87 million across 35 deals, giving the market an average disclosed round size of $26.54 million.
- Q1 2026 attracted $380 million, or 40.9% of all AI safety funding recorded during the five-quarter period.
- LMArena, Noma Security, and Braintrust raised $330 million together, equal to 35.5% of total AI safety market funding.
- AI risk platforms received about half of all disclosed capital because enterprises need one system for discovery, governance, access control, and monitoring.
- AI evaluation tools completed only three deals but raised $233.2 million, showing strong investor demand for trusted testing infrastructure.
- AI agents became the main safety concern in 2026, with funding focused on permissions, identities, tool access, runtime actions, and audit records.
- Q2 2026 had the broadest funding base, as the largest round represented only 22.4% of quarterly AI safety funding.
- Large seed rounds became common in the AI safety market, with JetStream Security, Guild.ai, Geordie AI, and NewCore each raising at least $30 million.
- Runtime controls attracted more funding than abstract safety research because enterprises can connect runtime protection to security and compliance budgets.
This market map, featured in our AI safety market deck, highlights top companies and startups in the AI safety market
Summary table of the funding deals in the AI safety market (last 5 quarters)
We define the AI safety market as products and services that reduce harms and failure modes from AI systems by measuring, mitigating, and governing model behavior and AI-specific risk.
We include AI evaluation and red teaming, safety monitoring and incident response, safety and guardrail layers, and AI trust, risk, and security management tools used to reduce risk.
We exclude generic MLOps, general cybersecurity, and general compliance or content moderation products unless the products directly address AI behavior or AI-specific threats.
You can also read our detailed analysis to understand how funding activity in the AI safety market has evolved over the last few years.
Also, you should know that we have a dedicated page, updated weekly, with all the latest fundraising deals in the AI safety market.
| Name | What they do | Amount in $ | Quarter |
|---|---|---|---|
| Virtue AI | Tests AI systems for security weaknesses and protects enterprise applications during production. | $30M | Q2 2025 |
| Pillar Security | Finds AI vulnerabilities during development and protects applications and agents after deployment. | $9M | Q2 2025 |
| Aurascape | Controls employee use of generative AI and prevents sensitive company data from leaking. | $50M | Q2 2025 |
| Unbound | Helps enterprises govern AI tools, route models, and prevent unsafe data sharing. | $4M | Q2 2025 |
| Repello AI | Continuously attacks generative AI applications and blocks the vulnerabilities it discovers. | $1.2M | Q2 2025 |
| Starseer | Maps enterprise AI risks and shows whether deployed models follow company policies. | $2M | Q3 2025 |
| Promptfoo | Tests AI applications for prompt injection, data leakage, harmful outputs, and reliability failures. | $18.4M | Q3 2025 |
| Noma Security | Discovers enterprise AI applications, tests weaknesses, and monitors agents during production. | $100M | Q3 2025 |
| ALIGNMT AI | Monitors healthcare AI for bias, safety, transparency, compliance, and operational risk. | $6.5M | Q3 2025 |
| Darwin AI | Helps government agencies build, document, deploy, and supervise AI systems responsibly. | $15M | Q4 2025 |
| Lumia Security | Discovers risky enterprise AI activity and applies security policies to users and agents. | $18M | Q4 2025 |
| LMArena | Compares leading AI models using open tests and large-scale human preference data. | $150M | Q1 2026 |
| WitnessAI | Controls how employees, applications, and autonomous agents interact with enterprise AI. | $58M | Q1 2026 |
| Coxwave | Analyzes conversational AI interactions to find failures and improve system reliability. | $4.8M | Q1 2026 |
| Braintrust | Helps teams evaluate, trace, monitor, and improve AI applications before and after deployment. | $80M | Q1 2026 |
| Portkey | Routes AI traffic while controlling reliability, costs, usage, security, and governance. | $15M | Q1 2026 |
| JetStream Security | Discovers enterprise AI assets and gives security teams governance and risk controls. | $34M | Q1 2026 |
| Guild.ai | Provides a controlled environment for deploying, governing, auditing, and reusing AI agents. | $30M | Q1 2026 |
| Galtea | Tests AI agents in realistic simulations before companies deploy the agents. | $3.2M | Q1 2026 |
| OpenBox AI | Verifies and documents autonomous AI-agent decisions for regulated and large enterprises. | $5M | Q1 2026 |
| Trent AI | Scans AI agents, code, infrastructure, and runtime behavior for security risks. | $13M | Q2 2026 |
| AIM Intelligence | Automatically attacks generative AI applications and applies defenses against discovered threats. | $7M | Q2 2026 |
| Capsule Security | Stops manipulated or misbehaving AI agents from completing unsafe actions. | $7M | Q2 2026 |
| Aigentsphere | Tracks enterprise AI-agent owners, permissions, actions, policies, and compliance status. | $4M | Q2 2026 |
| White Circle | Monitors AI outputs and intervenes when models drift or break company policies. | $11M | Q2 2026 |
| Tynapse | Blocks hallucinations, data leakage, and unsafe tool use by enterprise AI agents. | $3.17M | Q2 2026 |
| CodeIntegrity | Places permanent security boundaries around AI agents to prevent unpredictable actions. | $5M | Q2 2026 |
| Geordie AI | Controls how enterprise AI agents access identities, data, tools, and business systems. | $30M | Q2 2026 |
| Gray Swan | Tests frontier models and AI agents for harmful behavior and exploitable weaknesses. | $40M | Q2 2026 |
| Archestra.AI | Controls how AI agents connect to company data, software tools, and MCP servers. | $10M | Q2 2026 |
| ZeroDrift | Detects non-compliant AI communications and replaces them with approved alternatives. | $10M | Q2 2026 |
| Willow | Controls which systems, tools, and company data each AI agent can access. | $7M | Q2 2026 |
| Arcade.dev | Lets AI agents perform approved actions without receiving unrestricted company credentials. | $60M | Q2 2026 |
| NewCore | Provides secure corporate identities and access controls for people and autonomous agents. | $66M | Q2 2026 |
| NeuralTrust | Discovers, tests, monitors, and governs enterprise AI applications and autonomous agents. | $20M | Q2 2026 |
In our AI safety market deck, we identify pain points entrepreneurs should prioritize
How has funding activity in the AI safety market changed over time?
Q1 2026 was the largest quarter, mainly because LMArena, Braintrust, and WitnessAI raised a combined $288 million.
Q4 2025 was the smallest quarter because the AI safety market recorded only two disclosed rounds worth $33 million.
AI safety funding fell by 22.4% from Q1 2026 to Q2 2026, but Q2 2026 funding was 212.9% higher than Q2 2025.
After removing the largest one or two rounds from each quarter, the AI safety market still shows rising deal activity. Q2 2026 had more mid-sized rounds and relied less on a single company.
| Quarter | Number of deals | Total raised | Comment |
|---|---|---|---|
| Q2 2025 | 5 | $94M | Aurascape represented more than half of the quarter, while four smaller AI security rounds completed the market. |
| Q3 2025 | 4 | $127M | Noma Security contributed most of the quarter, making Q3 2025 highly dependent on one large deal. |
| Q4 2025 | 2 | $33M | Only Darwin AI and Lumia Security announced qualifying rounds, producing the quietest quarter in the dataset. |
| Q1 2026 | 9 | $380M | Three large rounds for LMArena, Braintrust, and WitnessAI pushed funding to its highest quarterly level. |
| Q2 2026 | 15 | $295M | Deal volume reached a record as investors backed many agent security and governance platforms. |
| All quarters | 35 | $929M | The AI safety market expanded from a small security niche into a broader enterprise infrastructure category. |
This chart, featured in our AI safety market deck, shows how HiddenLayer is positioned in AI safety
Which startups in the AI safety market raised the largest rounds over the last months?
These startups raised the most recently in the AI safety market:
- LMArena raised $150 million to build trusted model comparisons using open evaluations and large-scale human feedback.
- Noma Security raised $100 million as enterprises needed stronger discovery, governance, and runtime protection for AI applications and agents.
- Braintrust raised $80 million to become a central evaluation and observability layer for production AI applications.
- NewCore raised $66 million to build identity and access systems designed for both employees and autonomous AI agents.
- Arcade.dev raised $60 million to help enterprises authorize agent actions without exposing unrestricted credentials.
- WitnessAI raised $58 million to secure enterprise AI usage, expand internationally, and launch more controls for autonomous agents.
- Aurascape raised $50 million to scale its enterprise control layer for generative AI usage and data protection.
- Gray Swan raised $40 million to expand adversarial testing across frontier models, AI agents, and enterprise deployments.
- JetStream Security raised $34 million to launch an enterprise platform for discovering and governing generative and agentic AI systems.
- Virtue AI raised $30 million to commercialize a unified platform for AI red teaming, privacy, runtime protection, and security testing.
And, yes, we do cover most of them in our our beautiful pitch about the AI safety market.
You may also want to check our ranking of the most funded startups in the AI safety market as well as our list of the most valued startups.
This chart, featured in our AI safety market deck, shows annual funding in AI safety startups
Is the AI safety market shifting toward smaller or bigger deals?
The average disclosed AI safety funding round was $26.54 million across the full five-quarter period.
The quarterly average ranged from $16.50 million in Q4 2025 to $42.22 million in Q1 2026. Q1 2026 was higher because three companies raised more than $50 million.
Without the largest outliers, the AI safety market shows a gradual rise in mid-sized seed and Series A rounds rather than a clear move toward mega-deals.
| Quarter | Number of deals | Average deal size | Deals below $2M | Deals above $50M |
|---|---|---|---|---|
| Q2 2025 | 5 | $18.84M | 1 | 0 |
| Q3 2025 | 4 | $31.73M | 0 | 1 |
| Q4 2025 | 2 | $16.50M | 0 | 0 |
| Q1 2026 | 9 | $42.22M | 0 | 3 |
| Q2 2026 | 15 | $19.65M | 0 | 2 |
| All quarters | 35 | $26.54M | 1 | 6 |
This chart, featured in our AI safety market deck, compares the main business model options for AI alignment research labs
How concentrated was funding activity in the AI safety market?
AI safety funding was highly concentrated in Q3 2025, when Noma Security captured 78.8% of all capital. The AI safety market became less concentrated as deal volume increased in 2026.
Q2 2026 had the widest capital distribution, with the largest deal representing 22.4% of quarterly funding and the top three representing 56.3%.
| Quarter | Number of deals | Top 1 | Top 3 | Top 10 |
|---|---|---|---|---|
| Q2 2025 | 5 | 53.08% | 94.48% | 100.00% |
| Q3 2025 | 4 | 78.80% | 98.42% | 100.00% |
| Q4 2025 | 2 | 54.55% | 100.00% | 100.00% |
| Q1 2026 | 9 | 39.47% | 75.79% | 100.00% |
| Q2 2026 | 15 | 22.39% | 56.32% | 91.12% |
| All quarters | 35 | 16.15% | 35.53% | 71.76% |
This chart, featured in our AI safety market deck, shows revenue breakdown by customer segment in the AI safety market
Which categories in the AI safety market received the most funding?
AI risk platforms received $465.1 million, or 50.1% of total AI safety funding. Enterprises are buying broad platforms that combine discovery, governance, access controls, and runtime monitoring.
AI evaluation tools received $233.2 million, or 25.1% of total funding, from only three deals. LMArena and Braintrust showed that evaluation can become core AI infrastructure.
AI guardrail platforms received $126.17 million, or 13.6% of total funding under the stricter seven-deal classification. These products attract funding because companies need direct controls over AI inputs, outputs, and actions.
| Category name | Number of deals | Total raised | Startups and amount |
|---|---|---|---|
| AI Risk Platforms | 18 | $465.10M | Aurascape $50M; Unbound $4M; Starseer $2M; Noma Security $100M; ALIGNMT AI $6.5M; Darwin AI $15M; Lumia Security $18M; Portkey $15M; JetStream Security $34M; Guild.ai $30M; OpenBox AI $5M; Aigentsphere $4M; Geordie AI $30M; ZeroDrift $10M; Willow $7M; NewCore $66M; Arcade.dev $60M. |
| AI Evaluation Tools | 3 | $233.20M | LMArena $150M; Braintrust $80M; Galtea $3.2M. |
| AI Guardrail Platforms | 7 | $126.17M | Virtue AI $30M; Pillar Security $9M; WitnessAI $58M; Capsule Security $7M; White Circle $11M; Tynapse $3.17M; CodeIntegrity $5M. |
This chart, featured in our AI safety market deck, shows how prompt injection defense platform technology has evolved over time
Who are the biggest investors in the AI safety market?
Andreessen Horowitz was linked to at least three clearly disclosed AI safety deals, including Promptfoo, LMArena, and Braintrust, with ZeroDrift adding a possible fourth affiliated investment.
Samsung-affiliated investment entities participated in WitnessAI, AIM Intelligence, and Gray Swan, showing a strong interest in enterprise AI security and model testing.
Lightspeed Venture Partners participated in Virtue AI and LMArena, connecting AI application security with independent model evaluation.
Evolution Equity Partners backed Noma Security and NewCore, two large rounds focused on agent security and identity infrastructure.
Insight Partners invested in Promptfoo and Darwin AI, covering both adversarial AI testing and public-sector AI governance.
Mirae Asset Venture Investment backed AIM Intelligence and Tynapse, giving the investor exposure to South Korean AI red teaming and runtime guardrails.
Disclaimer: this investor list may be incomplete; we focus on publicly disclosed lead and prominent recurring investors, so some frequent minority participants may be underrepresented. “Total funded” does not represent the amount personally invested by an individual investor. Instead, it refers to the aggregate amount raised across all fundraising rounds in which the investor participated.
| Investor | Number of deals | Total funded | Startups |
|---|---|---|---|
| Andreessen Horowitz and affiliated vehicles | 3 to 4 | $248.4M to $258.4M | Promptfoo, LMArena, Braintrust, and possibly ZeroDrift under affiliated-vehicle treatment |
| Samsung-affiliated investment entities | 3 | $105M | WitnessAI, AIM Intelligence, Gray Swan |
| Lightspeed Venture Partners | 2 | $180M | Virtue AI, LMArena |
| Evolution Equity Partners | 2 | $166M | Noma Security, NewCore |
| Insight Partners | 2 | $33.4M | Promptfoo, Darwin AI |
| Mirae Asset group | 2 | $10.17M | AIM Intelligence, Tynapse |
This chart, featured in our AI safety market deck, shows how model risk management has driven growth in the AI safety market over time
Related blog posts
- All funding deals in the AI safety market
- Which startups have raised the most funding in the AI safety market?
- How strong is fundraising in the AI safety market right now?
