What are the fundraising trends in the AI safety market?
In our AI safety market deck, you will find everything you need to understand the market
SUMMARY
We analyzed every publicly disclosed equity round raised by pure-play AI safety companies between January 2024 and May 2026. We only kept equity rounds of $300K or more, excluded undisclosed-size financings, and focused on companies whose core activity is reducing harms and failure modes from AI systems by measuring, mitigating, or governing model behavior and AI-specific risk.
Across the full analyzed window, the AI safety market produced 54 qualifying disclosed rounds and about $1.43B of capital. The annual path is clearly upward: about $369M in 2024, about $626M in 2025, and about $439M already in year-to-date 2026.
The freshest signal is especially strong. Through May 2026, the AI safety market has already raised about $439M across 9 deals, versus about $96M across 4 deals over the comparable early-2025 period.
That growth is real, but it is not evenly distributed. The top 3 year-to-date 2026 rounds account for 81.5% of capital, and the smallest 4 deals account for only 3.9%, so the current market total mostly measures conviction in a few large infrastructure bets.
Round sizes have moved up sharply in 2026. The year-to-date median round is $30M and the average round is $48.8M, compared with a full-year 2025 median of $9M and average of $25M.
AI Evaluation Tools, Model Robustness Tools, and AI Safety Monitoring are the strongest capital magnets. Together, those categories captured about 89% of year-to-date 2026 capital, led by large rounds for LMArena, Goodfire, WitnessAI, and Fiddler AI.
AI Risk Platforms remain the most active category by deal count in 2026, with 3 of 9 deals. But they captured only 9.3% of capital, which suggests governance and risk tooling is still in company-formation mode while evaluation, interpretability, and monitoring are attracting larger conviction checks.
North America dominates the AI safety market by dollars. It represents 66.7% of year-to-date 2026 deals but 97.2% of capital, meaning non-North-American AI safety companies are visible but not yet capitalized at the same scale.
The market is shifting toward proven winners. Follow-on rounds account for 66.7% of year-to-date 2026 deals and 90.7% of capital, while first financings are still happening but carry only 9.3% of dollars.
Repeat investor activity is still shallow, but the quality of participation is rising. Lightspeed Venture Partners appears in three year-to-date 2026 deals, while Mozilla Ventures appears in two, showing stronger concentration around evaluation, monitoring, and model-understanding infrastructure.
This chart, featured in our AI safety market deck, shows revenue breakdown by customer segment in the AI safety market
Is more or less capital going into the AI safety market?
More capital is going into the AI safety market, but the answer needs a careful caveat: headline funding is growing while capital is becoming increasingly concentrated in a small number of large platform bets. Full-year funding rose from about $369M in 2024 to about $626M in 2025, and year-to-date 2026 has already reached about $439M.
The full-year comparison is the cleanest structural read. The AI safety market had 20 qualifying deals in 2024 and 25 in 2025, while capital grew much faster than deal count. That means the market did not merely add a few more companies; investors also wrote larger checks.
The freshest comparison is even more dramatic. Through May 2025, the AI safety market had 4 qualifying deals and about $96M of capital. Through May 2026, it had 9 deals and about $439M, so deal count more than doubled and capital rose more than fourfold.
The practical takeaway is that more capital is definitely entering the AI safety market, but the incremental dollars are not spread evenly across the category. LMArena and Goodfire each raised $150M, and WitnessAI raised $58M, so the current surge is mostly a signal of investor conviction in a few core infrastructure layers.
For deeper analysis of how capital is moving across the AI safety market, see the full AI safety market report.
Is AI safety funding driven by more deals or larger rounds?
AI safety funding is being driven by both more deals and larger rounds, but larger rounds explain more of the increase in capital. Deal count rose from 20 in 2024 to 25 in 2025, while total capital rose from about $369M to about $626M.
The median-versus-average split shows what is really happening. In 2024, the average round was about $18M and the median round was also about $18M, so the typical deal and the average deal looked similar. In 2025, the average rose to about $25M while the median fell to $9M, which means the market became more top-heavy.
Year-to-date 2026 is larger in the middle as well as at the top. The average round is about $49M and the median round is $30M, so the funding environment is not only being pulled upward by one outlier. The typical disclosed AI safety round in early 2026 is meaningfully larger than it was in the full-year 2025 dataset.
The honest interpretation is that the AI safety market has more activity than it had in 2024, but the funding surge is primarily a round-size story. Investors are paying substantially more when they believe a company can become a core measurement, monitoring, governance, or interpretability layer.
For more detail on medians, averages, deal-size buckets, and concentration ratios, see the AI safety market deck.
Is AI safety capital moving toward later-stage or earlier-stage companies?
AI safety capital is moving toward later-stage and more validated companies in the freshest 2026 period, even though early-stage company formation remains active. Year-to-date 2026 Series B+, Series C, and growth-equity rounds account for about 54% of capital.
The 2025 market looked earlier-stage on paper, with early-stage and unknown-stage rounds capturing about 84% of capital. But that period included several large rounds classified as unknown stage, so the stage label alone can understate how strategic some of those financings were.
The 2026 evidence is cleaner. Goodfire raised a $150M Series B, Fiddler AI raised a $30M Series C, and WitnessAI raised $58M in growth equity. Those are not exploratory seed checks; they are institutional commitments to companies that already had prior validation.
The AI safety market is therefore not abandoning new formation. Seed deals still represent 4 of 9 year-to-date 2026 deals. But the dollars have clearly shifted toward companies that investors already view as credible infrastructure providers.
This chart, featured in our AI safety market deck, compares the main business model options for AI alignment research labs
Is the AI safety market maturing or still experimental?
The AI safety market is maturing, but it is still not a fully mature late-stage venture market. The best read is that the market is moving from experimentation toward institutional validation, while many subcategories are still searching for durable packaging and buyer ownership.
The strongest maturity signal is the movement from first financings to follow-on capital. In 2025, first financings represented 68% of deals and about 54% of capital. In year-to-date 2026, first financings represent only 33% of deals and 9.3% of capital.
That shift matters because it shows investors are no longer only funding the idea that AI safety should exist. They are starting to select companies they believe can become durable platforms for evaluation, interpretability, monitoring, governance, or agent security.
The experimental side is still visible. AI Risk Platforms have the most 2026 deals but only 9.3% of capital, standalone guardrails have no qualifying pure-play deal in the year-to-date dataset, and red teaming has only one. Those are signs that parts of the AI safety market are still being defined.
The practical takeaway is that the AI safety market is no longer purely experimental. But its maturation is uneven: platform-like proof layers are attracting large checks, while narrower mitigation tools are still proving how they fit into enterprise budgets.
Are new startups still entering the AI safety market?
Yes, new startups are still entering the AI safety market, but the share of activity represented by first financings has slowed in the freshest 2026 data. In 2025, first financings represented 68% of deals and about 54% of capital; in year-to-date 2026, they represent 33% of deals and only 9.3% of capital.
The 2025 dataset is the strongest evidence that formation was healthy. The AI safety market produced 25 qualifying deals across 25 unique companies, with new entrants appearing across AI Risk Platforms, AI Evaluation Tools, AI Red Teaming, AI Guardrail Platforms, and Model Robustness Tools.
The 2026 picture is different but not negative. Skipr, JetStream Security, and OpenBox AI all entered the dataset as first financings, and all three sit in AI Risk Platforms. That suggests governance-grade visibility, trust infrastructure, auditability, and control layers are still open to new company creation.
The correct read is that the startup-formation frontier is narrowing. New companies can still enter the AI safety market, but the biggest checks are increasingly going to companies that already have category recognition or technical credibility.
For the broader view of new entrants and first financings in the AI safety market, see the deeper analysis of the AI safety market.
Are more investors entering the AI safety market?
More investors are showing up in the AI safety market in the freshest year-to-date comparison, even though the full-year 2024-to-2025 comparison is more mixed. Through May 2026, the dataset includes about 45 unique disclosed investors and 16 tier-1 investors across only 9 deals.
The comparable early-2025 window had only about 14 unique disclosed investors and 3 tier-1 investors. On that basis, the investor base is clearly broader and stronger early in 2026 than it was early in 2025.
The quality of the investor mix matters as much as the count. The 2026 dataset includes Lightspeed, Andreessen Horowitz, Kleiner Perkins, Felicis, Sound Ventures, Qualcomm Ventures, Samsung Ventures, Lux Capital, Insight Partners, Mozilla Ventures, B Capital, Menlo Ventures, Salesforce Ventures, DFJ Growth, Redpoint, and CrowdStrike Falcon Fund.
The better interpretation is that investor interest is becoming more selective but more serious. The AI safety market is attracting generalist venture firms, cybersecurity investors, enterprise software investors, and corporate strategics that see safety as an infrastructure adjacency.
This chart, featured in our AI safety market deck, shows annual funding in AI safety startups
Are top investors getting more or less active in AI safety?
Top investors are getting more active in AI safety, especially in the freshest 2026 period, but repeat behavior is still shallow compared with a fully mature venture category. In 2024, only YL Ventures, Hetz Ventures, and AI Fund appeared more than once; in 2025, the repeat list expanded to Insight Partners, Race Capital, Y Combinator, Mayfield, and Mirae Asset Capital.
Year-to-date 2026 has fewer repeat names, but the repeat signal is more important. Lightspeed appears in three deals: LMArena, Fiddler AI, and Goodfire. Those companies map directly to evaluation, monitoring, and interpretability/model robustness, which are three of the strongest capital themes in the AI safety market.
Mozilla Ventures also appears twice, in Fiddler AI and Galtea. That reinforces the same pattern: top investors are clustering around measurable, monitorable, and trustworthy AI infrastructure rather than generic responsible-AI language.
The honest interpretation is that no single investor syndicate controls the AI safety market yet. But the highest-quality investors are becoming more active where the infrastructure logic is clearest.
Which AI safety subcategories are gaining momentum?
AI Evaluation Tools, Model Robustness Tools, and AI Safety Monitoring are the AI safety subcategories gaining the most momentum. Together, they captured about 89% of year-to-date 2026 capital.
AI Evaluation Tools show the strongest continuity. The category captured about $103M in 2024, about $124M in 2025, and about $153M already in year-to-date 2026. LMArena’s $150M Series A is the main 2026 driver, but the broader pattern shows that investors increasingly treat evaluation as foundational AI infrastructure.
Model Robustness Tools are the clearest emerging category. The category had no qualifying capital in 2024, about $100M in 2025, and $150M in year-to-date 2026 from Goodfire alone. That suggests interpretability, model understanding, and controllable model behavior have moved from research-adjacent concepts into venture-scale commercial infrastructure.
AI Safety Monitoring is also gaining momentum, especially as enterprises move from AI experiments to production AI and agent deployment. WitnessAI and Fiddler AI show that runtime visibility, observability, policy enforcement, and governance telemetry are becoming fundable control layers.
We cover these subcategory shifts in more detail in the market report covering AI safety subcategories.
Which AI safety subcategories are losing momentum?
Standalone AI Guardrail Platforms and standalone AI Red Teaming are losing momentum as independent funding categories, even though guardrails and red teaming remain important capabilities inside broader AI safety platforms. The key point is about packaging, not usefulness.
AI Guardrail Platforms captured about $51M across 4 deals in 2024 and about $21M across 3 deals in 2025. In year-to-date 2026, no standalone pure-play guardrail-platform deal qualified. Guardrails are increasingly appearing inside monitoring, evaluation, governance, and red-teaming products instead of defining the whole company.
AI Red Teaming remains visible but is not where the largest pools of capital are going. It had one deal in 2024, four deals in 2025, and one deal in year-to-date 2026. The function is operationally important, but the market often treats it as a wedge or feature rather than a full platform category.
AI Risk Platforms are not losing momentum in deal count, but they are weaker in current capital share. They have the most 2026 deals but only 9.3% of capital, which suggests investors are still testing which governance architecture will actually win.
This chart, featured in our AI safety market deck, shows how HiddenLayer is positioned in AI safety
Which regions are gaining momentum in AI safety funding?
North America is gaining the most capital momentum in AI safety funding, while Europe and Asia-Pacific show narrower signs of company formation. The clearest signal is that North America captured 97.2% of year-to-date 2026 capital across 6 of 9 deals.
The full-year pattern was already moving in that direction. North America captured about 71% of 2024 AI safety capital and about 88% of 2025 capital. The 2026 year-to-date figure extends that trend and shows that the largest platform rounds are overwhelmingly North American.
Europe gained breadth in 2025, with 6 qualifying deals and about $56M of capital, up from 1 deal in 2024. In year-to-date 2026, Europe has one qualifying deal, Galtea, at $3.2M, which shows continuing formation but not large-scale funding participation.
Asia-Pacific is emerging from a very small base. It had no qualifying 2024 deals, two small deals in 2025, and one $7M AIM Intelligence round in year-to-date 2026. That is momentum, but still not enough to make the region a major AI safety capital center.
For ongoing regional tracking across North America, Europe, Asia-Pacific, the Middle East, and Latin America, see the full market view on AI safety funding geography.
Which regions are losing momentum in AI safety funding?
The Middle East is the clearest region losing momentum from its 2024 position, while Europe is not structurally weakening but is losing relative capital share in the freshest 2026 comparison. In 2024, the Middle East captured 6 deals and about $86M, or 23% of capital.
That 2024 strength did not continue at the same level. The 2025 regional split did not show Middle East activity among qualifying deals, and year-to-date 2026 includes one small Middle East deal, Skipr, at $2M.
Europe’s signal is more nuanced. Europe improved from 1 deal in 2024 to 6 deals in 2025, so it is not losing structural formation momentum. But it has not participated in the large early-2026 financing wave, which makes its capital share look very small next to North America.
The strongest regional conclusion is that the AI safety market is becoming more North-America-heavy in dollar terms. Other regions can still produce credible companies, but they are mostly operating at formation-stage scale.
Is AI safety becoming more global or regionally concentrated?
The AI safety market is becoming more global by company formation, but more regionally concentrated by capital. This distinction matters because the map of company creation and the map of large financing rounds are telling different stories.
In 2025, the AI safety market included North America, Europe, Latin America, and Asia-Pacific. That was more geographically diverse than the 2024 dataset, which was concentrated in North America, the Middle East, and Europe.
But the capital story moved the other way. North America captured about 88% of 2025 capital and 97.2% of year-to-date 2026 capital. That means non-North-American regions are contributing more logos than dollars.
The best interpretation is that AI safety is global at the seed layer and concentrated at the scale layer. Many geographies can create companies around regulation, evaluation, red teaming, governance, and sovereign AI trust, but the biggest rounds still depend on North American venture depth, enterprise buyers, AI infrastructure networks, and strategic investors.
This chart, featured in our AI safety market deck, shows how model risk management has driven growth in the AI safety market over time
Is AI safety capital moving toward proven winners or new opportunities?
AI safety capital is moving strongly toward proven winners in 2026, even though new opportunities remain funded at the seed layer. The clearest indicator is that follow-on rounds represent 66.7% of year-to-date 2026 deals and 90.7% of capital.
The company-level evidence is blunt. LMArena, WitnessAI, Fiddler AI, Goodfire, Galtea, and AIM Intelligence were all follow-on rounds in year-to-date 2026. Together, they captured nearly all of the current-year capital.
New opportunities still matter, but they are mostly smaller and clustered in AI Risk Platforms. Skipr raised $2M, JetStream Security raised $34M, and OpenBox AI raised $5M as first financings. JetStream is the exception that proves new companies can still raise large rounds when the problem framing and investor syndicate are strong.
The strongest conclusion is that the AI safety market is shifting from a formation market to a selection market. Investors are increasingly asking which companies have enough proof to become the control plane, measurement layer, interpretability layer, or monitoring system of record.
Our AI safety market report tracks which companies are attracting follow-on capital and which new entrants still need to prove they can raise again.
Is the AI safety market becoming winner-takes-most?
The AI safety market is becoming more winner-takes-most in capital allocation, though not yet necessarily in product adoption. The strongest evidence is concentration: the top 3 deals captured 38.2% of 2024 capital, 48.0% of 2025 capital, and 81.5% of year-to-date 2026 capital.
The bottom-half share confirms the pattern. In 2024, the bottom half of deals captured 19.7% of capital. In 2025, the bottom half captured 7.5%. In year-to-date 2026, the smallest 4 deals captured only 3.9%.
That does not automatically prove the product market is winner-takes-most. Funding concentration is not the same thing as revenue concentration, customer concentration, or category dominance. It proves that investor conviction is concentrating faster than company count.
The best interpretation is that AI safety is becoming winner-takes-most in financing narratives. Investors are willing to fund perceived category leaders at large scale while leaving many smaller startups with modest seed or early-stage rounds.
Is the next wave of AI safety winners becoming visible?
Yes, the next wave of AI safety winners is becoming visible, especially in evaluation, interpretability/model robustness, monitoring, and enterprise AI control planes. The largest year-to-date 2026 rounds are not random; they cluster around functions investors believe will matter most.
LMArena raised $150M in AI Evaluation Tools, Goodfire raised $150M in Model Robustness Tools, WitnessAI raised $58M in AI Safety Monitoring, and Fiddler AI raised $30M in monitoring and governance. Those financings point toward companies that can make AI behavior measurable, governable, auditable, monitorable, and steerable.
The next wave is not fully settled. AI Risk Platforms produced the most 2026 deals but only 9.3% of capital, which means investors still believe new governance and control-plane opportunities exist but have not crowned one clear winner there.
The practical filter is simple. Companies that produce recurring proof, such as benchmarks, telemetry, audit trails, policy enforcement records, model-behavior explanations, or runtime risk signals, look more durable than companies that only describe safety in broad ethical or compliance language.
As this chart shows, and as featured in our AI safety market deck, search interest in AI safety has been growing steadily
Is the AI safety funding landscape fragmenting or consolidating?
The AI safety funding landscape is consolidating in dollars but fragmenting in company formation and product packaging. Capital is concentrating sharply, while the set of active workflows remains spread across evaluation, monitoring, risk platforms, red teaming, model robustness, and guardrail-like features.
The consolidation signal is obvious at the top of the capital stack. The top 10 deals captured 80.3% of 2024 capital and 87.2% of 2025 capital, while the top 5 deals captured 96.1% of year-to-date 2026 capital.
The fragmentation signal is visible in subcategories. In 2025, all six tracked categories received qualifying capital. In 2026, guardrails disappear as a standalone category, but guardrail functionality still appears inside red-teaming, monitoring, governance, and risk-platform products.
The right way to describe the AI safety market is asymmetric. Funding is consolidating around a few perceived platforms, while product boundaries are fragmenting and blurring as companies combine measurement, enforcement, observability, governance, and model control.
Where is investor attention shifting in AI safety?
Investor attention in the AI safety market is shifting toward measurable, auditable, production-grade AI trust infrastructure. The winning narratives increasingly sound less like broad responsible-AI statements and more like deployment systems: evaluate models, observe agents, enforce policies, harden behavior, and produce audit evidence.
The first shift is toward evaluation as a system of proof. LMArena, Galtea, and earlier evaluation companies show that investors want tools that can measure whether models and agents actually behave reliably before and after deployment.
The second shift is toward interpretability and model robustness. Goodfire’s $150M round is important because it shows that model understanding, editing, monitoring, and controllable behavior are now considered commercial infrastructure, not only research topics.
The third shift is toward runtime control. WitnessAI, Fiddler AI, OpenBox AI, and JetStream Security all point toward the same buyer need: enterprises want visibility, policy enforcement, auditability, and control as AI systems move into production.
For real-time tracking of how investor attention is moving across evaluation, red teaming, guardrails, monitoring, risk platforms, and model robustness, see the AI safety market report.
All the funding deals in the AI safety market from 2024 to Apr 2026
The table below lists every disclosed equity round in the supplied AI safety, AI governance, AI risk, AI evaluation, guardrails, red-teaming, monitoring, and robustness dataset from January 2024 through April 2026.
Each row shows the company, the fundraising date, what the company does, its category, the funding stage, the round size, the region, whether it was a first financing or a follow-on, the tier-1 investor if any, and the announcement source. For the broader investability view, see our market deck.
| Company | Date | What they do | Category | Stage | Deal size | Region | First/Follow-on | Tier 1 investor(s) | Source |
|---|---|---|---|---|---|---|---|---|---|
| AIM Intelligence | Apr 2026 | AI security platform combining automated AI red teaming and guardrails for LLMs, AI agents, and multimodal systems. | AI Red Teaming | Series A | $7M | Asia-Pacific | Follow-on | Samsung Venture Investment | Wowtale |
| OpenBox AI | Mar 2026 | Enterprise AI trust and governance platform for runtime policy enforcement, audit trails, cryptographic attestation, human oversight, and agent risk scoring. | AI Risk Platforms | Seed | $5M | North America | First financing | None identified | OpenBox AI |
| Galtea | Mar 2026 | AI evaluation infrastructure for generating test scenarios and evaluating agent performance, accuracy, security, and robustness before deployment. | AI Evaluation Tools | Seed | $3.2M | Europe | Follow-on | Mozilla Ventures | Tech.eu |
| JetStream Security | Mar 2026 | AI governance and control layer for enterprise AI systems, focused on visibility, risk management, trust, and production readiness. | AI Risk Platforms | Seed | $34M | North America | First financing | Redpoint Ventures; CrowdStrike Falcon Fund | JetStream Security |
| Skipr | Feb 2026 | Autonomous trust fabric for AI systems, using cryptographic identity, policy routing, and auditable interoperability for sovereign AI deployments. | AI Risk Platforms | Seed | $2M | Middle East | First financing | None disclosed | Hub71 |
| Goodfire | Feb 2026 | AI interpretability platform and research lab for understanding, editing, monitoring, and intentionally designing model behavior. | Model Robustness Tools | Series B | $150M | North America | Follow-on | B Capital; Menlo Ventures; Lightspeed Venture Partners; Salesforce Ventures; DFJ Growth | Goodfire |
| Fiddler AI | Jan 2026 | AI observability, security, evaluation, monitoring, policy, and auditable governance control plane for compound AI and agents. | AI Safety Monitoring | Series C | $30M | North America | Follow-on | Lightspeed Venture Partners; Lux Capital; Insight Partners; Mozilla Ventures | Fiddler AI |
| WitnessAI | Jan 2026 | Enterprise AI security and governance platform for observing, controlling, and protecting AI and AI-agent activity. | AI Safety Monitoring | Growth Equity | $58M | North America | Follow-on | Sound Ventures; Qualcomm Ventures; Samsung Ventures | WitnessAI |
| LMArena | Jan 2026 | Community and enterprise platform for real-world, reproducible AI model evaluation. | AI Evaluation Tools | Series A | $150M | North America | Follow-on | Andreessen Horowitz; Kleiner Perkins; Lightspeed Venture Partners; Felicis | LMArena |
| Mirror Security | Dec 2025 | AI security platform combining encryption, automated red teaming, and agentic security. | Model Robustness Tools | Seed | $2.5M | Europe | First financing | None clearly tier-1 | Mirror Security |
| Alinia AI | Dec 2025 | Guardrails API and AI compliance platform for regulated AI systems, especially financial AI agents. | AI Guardrail Platforms | Seed | $7.5M | Europe | First financing | Speedinvest | Tech Funding News |
| Vijil | Nov 2025 | Trust, governance, resilience, monitoring, and defense infrastructure for enterprise AI agents. | Model Robustness Tools | Series A | $17M | North America | Follow-on | Mayfield | SiliconANGLE |
| Portal26 | Nov 2025 | GenAI governance, security, and analytics platform for responsible enterprise AI adoption. | AI Risk Platforms | Series A | $9M | North America | Follow-on | Shasta Ventures | Tech Funding News |
| AI Score | Nov 2025 | Centralized AI governance and management platform for visibility, compliance monitoring, and risk scoring. | AI Risk Platforms | Unknown | $1M | Europe | First financing | None identified | StartupMag |
| Polygraf AI | Oct 2025 | Secure AI platform for enterprise, defense, and intelligence use cases, using trusted small language models. | AI Risk Platforms | Seed | $9.5M | North America | First financing | None clearly tier-1 | Business Wire |
| Darwin AI | Oct 2025 | AI governance and visibility platform for public-sector AI deployments. | AI Risk Platforms | Series A | $15M | North America | Follow-on | Insight Partners | PR Newswire |
| nexos.ai | Sep 2025 | Enterprise AI orchestration/governance platform for secure AI adoption, model routing, and control. | AI Risk Platforms | Series A | $35M | Europe | Follow-on | Index Ventures | TechCrunch |
| Scorecard | Sep 2025 | AI agent evaluation and testing platform for trusted deployment. | AI Evaluation Tools | Seed | $3.75M | North America | First financing | None identified | Scorecard |
| Trismik | Sep 2025 | Science-grade LLM evaluation startup measuring AI capabilities. | AI Evaluation Tools | Seed | $2.9M | Europe | First financing | None clearly tier-1 | UK Tech News |
| Irregular | Sep 2025 | Frontier AI model security lab testing models for cyber misuse and resilience. | Model Robustness Tools | Unknown | $80M | North America | Follow-on | Sequoia Capital; Redpoint Ventures | TechCrunch |
| Airia | Sep 2025 | Enterprise AI security, governance, and orchestration platform for safer AI adoption. | AI Risk Platforms | Unknown | $100M | North America | First financing | None; founder-backed | SiliconANGLE |
| Podonos | Sep 2025 | Infrastructure for evaluating voice AI model quality, reliability, and performance. | AI Evaluation Tools | Seed | $2.4M | North America | First financing | None identified | Podonos |
| AIM Intelligence | Aug 2025 | Generative AI security platform for automated red teaming, guardrails, and controllability. | AI Red Teaming | Unknown | $1.3M | Asia-Pacific | First financing | None clearly tier-1 | KoreaTechDesk |
| Noma Security | Jul 2025 | AI agent security platform for monitoring, threat detection, testing, and safeguards. | AI Safety Monitoring | Series B | $100M | North America | Follow-on | Evolution Equity Partners; Databricks Ventures | Wall Street Journal |
| Promptfoo | Jul 2025 | Open-source AI security testing, red-teaming, evaluation, guardrails, and continuous monitoring platform. | AI Red Teaming | Series A | $18.4M | North America | Follow-on | Insight Partners; Andreessen Horowitz | SiliconANGLE |
| Repello AI | Jun 2025 | GenAI security platform for automated red teaming and runtime protection. | AI Red Teaming | Seed | $1.2M | Asia-Pacific | First financing | None clearly tier-1 | Repello AI |
| Trustible | Jun 2025 | Enterprise AI governance platform for responsible AI adoption, compliance, and risk workflows. | AI Risk Platforms | Seed | $4.6M | North America | First financing | None clearly tier-1 | Technical.ly |
| Unbound | May 2025 | AI gateway giving enterprises visibility, protection, routing, and governance over AI tool use. | AI Guardrail Platforms | Seed | $4M | North America | First financing | Y Combinator | Yahoo Finance |
| LMArena | May 2025 | Open AI model evaluation platform focused on reliable measurement and human preference benchmarking. | AI Evaluation Tools | Seed | $100M | North America | First financing | a16z; Lightspeed; Kleiner Perkins | PR Newswire |
| Openlayer | May 2025 | AI testing and governance platform for pre- and post-deployment evaluation. | AI Evaluation Tools | Series A | $14.5M | Latin America | Follow-on | Y Combinator | LatamList |
| Pillar Security | Apr 2025 | AI application and infrastructure security platform with AI lifecycle guardrails. | AI Guardrail Platforms | Seed | $9M | North America | First financing | None clearly tier-1 | Refresh Miami |
| Virtue AI | Apr 2025 | Unified AI security, red-teaming, compliance, and risk platform for enterprise AI systems. | AI Risk Platforms | Unknown | $30M | North America | First financing | Lightspeed Venture Partners | Business Wire |
| SplxAI | Apr 2025 | AI security and red-teaming platform for identifying vulnerabilities in AI systems. | AI Red Teaming | Seed | $7M | Europe | First financing | None clearly tier-1 | Business Insider |
| Aurascape | Mar 2025 | Enterprise platform to detect and mitigate shadow AI risk and secure AI usage. | AI Risk Platforms | Unknown | $50M | North America | First financing | Menlo Ventures; Mayfield Fund | SecurityWeek |
| Fiddler AI | Dec 2024 | AI observability, model monitoring, explainability, and AI safety platform. | AI Safety Monitoring | Series B | $18.6M | North America | Follow-on | None clearly identified | Dallas Innovates |
| Prompt Security | Nov 2024 | GenAI security platform for employees, applications, and customers. | AI Guardrail Platforms | Series A | $18M | Middle East | Follow-on | Okta; F5; Jump Capital | Prompt Security |
| Noma Security | Oct 2024 | Platform securing the data and lifecycle of GenAI applications against new AI-specific threats. | AI Safety Monitoring | Series A | $32M | Middle East | Follow-on | Ballistic Ventures; Glilot Capital Partners | Yahoo Finance |
| Galileo | Oct 2024 | Generative AI evaluation and observability platform / Evaluation Intelligence for trustworthy AI. | AI Evaluation Tools | Series B | $45M | North America | Follow-on | Scale Venture Partners; Databricks Ventures; Citi Ventures; ServiceNow; SentinelOne | Galileo |
| Braintrust | Oct 2024 | AI evaluations and observability platform for building, testing, and improving AI products. | AI Evaluation Tools | Series A | $36M | North America | Follow-on | Andreessen Horowitz; Greylock; Datadog; Databricks Ventures; Elad Gil | Braintrust |
| ModelOp | Aug 2024 | Enterprise AI governance software for operationalizing governance across AI lifecycle. | AI Risk Platforms | Series B | $10M | North America | Follow-on | Baird Capital | ModelOp |
| Protect AI | Aug 2024 | AI/ML security platform for AI security posture management, visibility, remediation, control, and governance of AI/ML systems. | AI Risk Platforms | Series B | $60M | North America | Follow-on | Evolution Equity Partners; Salesforce Ventures; Samsung; StepStone; Acrew; Boldstart | Business Wire |
| Credo AI | Jul 2024 | AI governance software for responsible AI use, compliance, and AI risk management. | AI Risk Platforms | Series B | $21M | North America | Follow-on | Mozilla Ventures; Sands Capital; Decibel VC; AI Fund; Booz Allen Hamilton | Business Wire |
| Lakera | Jul 2024 | Real-time GenAI security platform including Lakera Guard and Gandalf red-team dataset/product. | AI Guardrail Platforms | Series A | $20M | Europe | Follow-on | Atomico; Citi Ventures; Dropbox Ventures | Lakera |
| Aim Security | Jun 2024 | GenAI enterprise security platform. | AI Risk Platforms | Series A | $18M | Middle East | Follow-on | Canaan Partners; YL Ventures | Business Wire |
| Patronus AI | May 2024 | Automated evaluation and security platform for detecting LLM mistakes and supporting safe enterprise GenAI deployment. | AI Evaluation Tools | Series A | $17M | North America | Follow-on | Lightspeed Venture Partners; Datadog; Notable Capital | Patronus AI |
| WitnessAI | May 2024 | Secure AI enablement platform providing visibility, control, and protection for enterprise AI use. | AI Safety Monitoring | Series A | $27.5M | North America | First financing | GV | PR Newswire |
| Knostic | Apr 2024 | Need-to-know access controls for enterprise GenAI systems to reduce oversharing/data-exposure risk. | AI Risk Platforms | Unknown | $3.3M | Middle East | First financing | Seedcamp; Shield Capital | CTech |
| SydeLabs | Mar 2024 | GenAI security and risk management platform, including red-teaming capabilities. | AI Red Teaming | Seed | $2.5M | North America | First financing | RTP Global | SecurityWeek |
| ValidMind | Mar 2024 | AI and model risk management/governance platform for financial institutions. | AI Risk Platforms | Seed | $8.1M | North America | Follow-on | Point72 Ventures; AI Fund; Notion Capital; FJ Labs | Business Wire |
| Guardrails AI | Mar 2024 | Open-source validation and correction framework / AI assurance layer for trustworthy LLM outputs. | AI Guardrail Platforms | Seed | $7.5M | North America | First financing | Bloomberg Beta; Pear VC; GitHub Fund; Zetta | Zetta Venture Partners |
| Armilla AI | Feb 2024 | AI assurance and warranty/risk platform for validating AI products and mitigating enterprise AI risk. | AI Risk Platforms | Seed | $4.5M | North America | First financing | Y Combinator | SME Business Review |
| Aim Security | Jan 2024 | GenAI security platform for secure enterprise AI adoption and unique GenAI risks. | AI Risk Platforms | Seed | $10M | Middle East | First financing | YL Ventures | VentureBeat |
| Prompt Security | Jan 2024 | Enterprise GenAI security platform preventing sensitive-data leakage and securing AI products from prompt injection and jailbreak risks. | AI Guardrail Platforms | Seed | $5M | Middle East | First financing | None clearly identified | Prompt Security |
| RagaAI | Jan 2024 | AI testing platform using RagaAI DNA to test AI products and services for model/data failures. | AI Evaluation Tools | Seed | $4.7M | North America | First financing | None clearly identified | Entrackr |
INSIGHTS
The insights below come from reviewing every disclosed equity round in the AI safety market between January 2024 and May 2026, including the full-year 2024 dataset, the full-year 2025 dataset, and the year-to-date 2026 dataset.
- The AI safety market’s headline growth is real, but growth is increasingly explained by capital concentration rather than evenly distributed health. Funding rose from about $369M in 2024 to about $626M in 2025 and about $439M through May 2026, but the top 3 year-to-date 2026 deals explain more than 80% of current-year capital.
- The market has moved from proving that AI safety matters to selecting which infrastructure layers will own the budget. Evaluation, interpretability, monitoring, governance, and control-plane companies are getting the clearest funding signals.
- The 2025 market looked broader, while the 2026 market looks more decisive. In 2025, 25 companies raised across many categories; in early 2026, only 9 companies raised, but the largest rounds were bigger and more concentrated around platform infrastructure.
- Average round size is becoming less representative of the typical AI safety company. In 2025, the average round was about $25M while the median was $9M, and in year-to-date 2026 the average is about $49M while the median is $30M. The gap shows how easily a few large rounds can distort the funding climate.
- The rise in the median round to $30M in early 2026 is more important than the rise in the average round. A high average can be created by one outlier, but a high median means the middle of the disclosed deal set is also moving toward larger institutional checks.
- The decline in first-financing capital share from about 54% in 2025 to about 9% in early 2026 is one of the clearest signs of category selection. Investors are still funding new startups, but they are reserving most dollars for companies with prior validation.
- Standalone guardrails are losing their ability to define a venture-scale category by themselves. The absence of a qualifying standalone guardrail-platform deal in year-to-date 2026 suggests buyers increasingly expect guardrails to be bundled inside broader monitoring, red-teaming, evaluation, or governance platforms.
- Red teaming is more important operationally than financially. The function appears across many company narratives, but standalone red-teaming companies capture relatively small capital shares, suggesting red teaming is often a wedge, feature, or service layer rather than the final platform.
- AI Evaluation Tools are becoming a market anchor because they provide measurable proof. Investors increasingly treat evaluation as foundational AI infrastructure, especially when it can support benchmarks, deployment gates, and ongoing model or agent assessment.
- Model Robustness Tools moved from absent in 2024 to meaningful in 2025 and central in early 2026. That progression suggests interpretability and model-behavior hardening have crossed from research-adjacent topics into venture-scale commercial infrastructure.
- AI Safety Monitoring is becoming a validation category for agent deployment. WitnessAI, Noma Security, and Fiddler AI show that investors are willing to fund runtime visibility, control, and observability when AI systems move from experiments into production.
- AI Risk Platforms are broad in deal count but weaker in current capital share. The category has many new entrants because the need for governance is obvious, but the market has not yet resolved which governance architecture will win.
- Capital is shifting from static assurance to continuous control. The strongest company narratives emphasize live monitoring, policy enforcement, auditable telemetry, agent oversight, model behavior evaluation, and production governance rather than one-time compliance checks.
- The AI safety market is more North-American in dollars than in ideas. Europe, Asia-Pacific, Latin America, and the Middle East all produce qualifying companies, but North America captures nearly all of the largest financing rounds.
- The investor base is broad but not deeply syndicated. A few investors repeat across deals, but there is not yet a stable group of specialist AI safety investors that dominates the market.
- Lightspeed’s 2026 pattern is unusually important because it spans LMArena, Fiddler AI, and Goodfire. That cross-category participation suggests a thesis around measurable, monitorable, and interpretable AI behavior rather than a narrow bet on one subcategory.
- Strategic and corporate investors matter in this market. Qualcomm Ventures, Samsung Ventures, Salesforce Ventures, Databricks Ventures, CrowdStrike Falcon Fund, and similar participants validate that AI safety is becoming an infrastructure adjacency for security, cloud, data, and enterprise platforms.
- The market rewards companies that attach safety to a hard operational bottleneck. The strongest funding signals go to companies solving evaluation, agent monitoring, governance-grade visibility, interpretability, auditability, and production control, not companies with vague responsible-AI messaging.
- The market’s category boundaries are blurring. Evaluation platforms add monitoring, monitoring platforms add governance, governance platforms add risk scoring, red-teaming platforms add guardrails, and robustness tools add interpretability and model control.
- Category blurring favors platform companies over point tools. Companies that combine measurement, enforcement, observability, and governance are more likely to attract large checks than companies solving one narrow safety workflow.
- The concentration of capital does not yet prove product-market concentration. Investors are anointing perceived winners, but customer adoption, retention, integration depth, and budget ownership will determine whether financing concentration turns into market concentration.
- The most useful diligence rule is to ask what evidence the product produces. Products that generate benchmarks, audit trails, telemetry, policy enforcement records, model-behavior explanations, or runtime risk signals deserve more weight than products that only describe governance aspirations.
- The biggest unanswered question is buyer ownership. AI safety can be sold to security, engineering, compliance, risk, data, AI platform, or executive buyers, and the companies that resolve buyer ownership cleanly will likely outperform companies trapped between departments.
This chart, featured in our AI safety market deck, shows how prompt injection defense platform technology has evolved over time
OUR METHODOLOGY TO BUILD THIS TRACKER
We built this AI safety funding tracker by reviewing every publicly disclosed equity round raised by pure-play AI safety companies between January 2024 and May 2026. A company counts as pure-play when more than 80% of its activity is dedicated to reducing harms and failure modes from AI systems by measuring, mitigating, monitoring, hardening, or governing model behavior and AI-specific risk.
We applied four filters to build the dataset. First, we only included equity rounds, so grants, debt, structured financings, SPAC transactions, acquisitions, and business combinations were excluded. Second, we only counted rounds of $300K or more. Third, we only kept pure-play AI safety companies, which means we excluded generic MLOps, generic cybersecurity, generic GRC, frontier model labs, and AI application companies where AI safety was not the core product. Fourth, every entry had to be confirmed by a direct company announcement, press release, tier-1 media report, specialized industry source, or relevant regional publication.
The included categories are AI Evaluation Tools, AI Red Teaming, AI Guardrail Platforms, AI Safety Monitoring, AI Risk Platforms, and Model Robustness Tools. We used the announcement month and year as the fundraising date unless the source clearly indicated a different close timing. Undisclosed-amount rounds were excluded because including them would distort dollar-based metrics such as averages, medians, category capital shares, and concentration ratios.
The final disclosed dataset covers 20 qualifying rounds in 2024, 25 in 2025, and 9 in year-to-date 2026. Privately raised rounds, rounds hidden behind paid databases, and rounds announced without reliable public amounts are necessarily missing, which is a known limitation of any public-only AI safety funding tracker.
Related blog posts
- The latest update in the AI safety market
- The latest news in the AI safety market
- How big is the AI safety market really?
- The latest funding news in the AI safety market
- The evolution of funding activity in the AI safety market
- Which startups have raised the most funding in the AI safety market?
- Which startups are the most valued in the AI safety market?
