AI Safety Startup Funding 2025-2026
In our AI safety market deck, you will find everything you need to understand the market
SUMMARY
We have analyzed every publicly disclosed equity round raised by pure-play or near-pure-play AI safety companies between July 2025 and June 2026, a 12-month window covering disclosed deals available through June 8, 2026. We only kept equity rounds of $300K or more, and excluded generic MLOps, general cybersecurity, broad compliance, and content moderation companies unless the product was specifically designed to reduce AI-model behavior risk or AI-specific threats.
Over this period, fundraising in the AI safety market was active but narrow. The dataset includes 20 disclosed deals, 20 unique companies, and $560.27M of disclosed capital raised.
The AI safety market is heavily shaped by a small number of large rounds. The top deal alone represents 26.77% of total capital, the top 3 deals reach 58.90%, and the top 5 deals reach 75.32%.
The typical AI safety round is much smaller than the headline average suggests. The median round size is $9.00M, while the average round size is $28.01M because LMArena, Noma Security, Braintrust, and WitnessAI pull the mean upward.
Deal flow averaged 1.67 rounds per month over the 12-month period. Capital raised averaged $46.69M per month, but that number is highly sensitive to a few months with very large Series A and Series B rounds.
AI Evaluation Tools leads the AI safety market by dollars, with $230.00M raised and 41.05% of total capital. AI Risk Platforms lead by deal count, with 9 deals and 45.00% of disclosed activity.
North America dominates the AI safety market by capital. The region accounts for $491.90M, or 87.80% of total disclosed funding, despite representing only 60.00% of deals.
The market is still mostly early-stage by deal count. Seed rounds represent 10 of 20 deals, but Series A and Series B rounds together capture 81.82% of total capital.
Early-stage rounds, defined here as Seed, Series A, and Unknown, represent $322.27M, or 57.52% of disclosed capital. Late-stage rounds, defined here as Series B and later, represent $238.00M, or 42.48%.
Repeat investor activity is limited but meaningful. Andreessen Horowitz appears in 3 deals, while Insight Partners and Mirae Asset group each appear in 2.
This market map, featured in our AI safety market deck, highlights top companies and startups in the AI safety market
What are all the funding deals in the AI safety market from July 2025 to June 2026?
The table below lists every disclosed equity round raised by pure-play or near-pure-play AI safety companies between July 2025 and June 2026. 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.
Each row shows the company, what it does, its category, the deal date, the funding stage, the round size, the region, the main investors, and the announcement source. For a wider view of how AI evaluation, red teaming, guardrails, monitoring, and AI risk platforms fit together, we cover it in our AI Safety market report.
| Company | What they do | Category | Date | Stage | Deal size | Region | Main investors | Source |
|---|---|---|---|---|---|---|---|---|
| Starseer | AI security, transparency, and reliability technology for making AI systems more understandable and secure | AI Risk Platforms | Jul 2025 | Seed | $2M | North America | Gula Tech Adventures | Business Wire |
| Promptfoo | Open-source and enterprise AI red-teaming, security testing, and evaluation framework for LLM applications | AI Red Teaming | Jul 2025 | Series A | $18.4M | North America | Insight Partners; Andreessen Horowitz | Promptfoo |
| Noma Security | Unified AI and AI-agent security platform for monitoring, testing, governing, and defending enterprise AI agents | AI Safety Monitoring | Jul 2025 | Series B | $100M | North America | Evolution Equity Partners | Noma Security |
| ALIGNMT AI | Healthcare AI compliance, governance, and risk-monitoring platform for safe, transparent, and equitable AI systems | AI Risk Platforms | Sep 2025 | Seed | $6.5M | North America | AIX Ventures | ALIGNMT AI |
| eRoun&Company | Generative AI security and AI governance company, including SAIFE X for AI governance and security use cases | AI Guardrail Platforms | Oct 2025 | Unknown | $1.4M | Asia-Pacific | KB Investment | Wowtale |
| Darwin AI | AI governance and workflow platform for governments deploying AI safely, transparently, and accountably | AI Risk Platforms | Oct 2025 | Series A | $15M | North America | Insight Partners | PR Newswire |
| Lumia Security | Agentic AI security and governance platform for visibility, control, policy enforcement, and guardrails | AI Risk Platforms | Dec 2025 | Seed | $18M | North America | Team8 | Lumia Security |
| LMArena | Community and enterprise AI evaluation platform for measuring model performance, reliability, and human preference signals | AI Evaluation Tools | Jan 2026 | Series A | $150M | North America | Felicis; UC Investments; Andreessen Horowitz | PR Newswire |
| WitnessAI | Enterprise AI security, governance, and guardrail platform for securing AI applications, LLM interactions, and AI agents | AI Guardrail Platforms | Jan 2026 | Series B | $58M | North America | Sound Ventures; Qualcomm; Samsung | Axios |
| Coxwave | AI trust technology and agent verification platform with reliability and governance capabilities for AI systems | AI Safety Monitoring | Jan 2026 | Unknown | $4.8M | Asia-Pacific | Undisclosed | Wowtale |
| Braintrust | AI evaluation and observability platform for testing, tracing, monitoring, and improving production AI systems and agents | AI Evaluation Tools | Feb 2026 | Series B | $80M | North America | ICONIQ; Andreessen Horowitz | Axios |
| JetStream Security | Enterprise AI governance and control platform providing visibility, blueprints, and risk controls for production AI usage | AI Risk Platforms | Mar 2026 | Seed | $34M | North America | Redpoint Ventures; CrowdStrike Falcon Fund | JetStream Security |
| OpenBox AI | Enterprise AI trust and governance infrastructure for oversight, verification, policy enforcement, and auditability of AI agents | AI Risk Platforms | Mar 2026 | Seed | $5M | North America | Tykhe Ventures | OpenBox AI |
| AIM Intelligence | Enterprise generative-AI security platform covering automated AI red teaming and real-time guardrails | AI Red Teaming | Apr 2026 | Series A | $7M | Asia-Pacific | Samsung Venture Investment; Mirae Asset group | Seoul Economic Daily |
| Tynapse | Runtime AI-agent security and trust layer for blocking hallucinations, data leakage, jailbreaks, and unsafe agent actions | AI Guardrail Platforms | Apr 2026 | Seed | $3.17M | Asia-Pacific | Mirae Asset group | Wowtale |
| Aigentsphere | Governance infrastructure for enterprise autonomous AI agents, including monitoring, policy enforcement, and compliance reporting | AI Risk Platforms | Apr 2026 | Seed | $4M | Asia-Pacific | Main Sequence | Startup Daily |
| White Circle | AI control platform that monitors, protects, tests, and improves AI models in production | AI Guardrail Platforms | May 2026 | Seed | $11M | Europe | Industry investors | Business Wire |
| CodeIntegrity | Security guardrails for agentic AI applications focused on prompt injection and non-deterministic AI-agent flaws | AI Guardrail Platforms | May 2026 | Seed | $5M | North America | Undisclosed | GeekWire |
| Geordie AI | Security and governance platform for enterprise AI agents, providing visibility, control, and safe adoption infrastructure | AI Risk Platforms | May 2026 | Series A | $30M | Europe | Balderton Capital | Geordie AI |
| Willow | Agentic access, identity, oversight, and governance platform for securing enterprise autonomous AI agents | AI Risk Platforms | Jun 2026 | Seed | $7M | Middle East | Hetz Ventures | PR Newswire |
In our AI safety market deck, we identify pain points entrepreneurs should prioritize
OUR METHODOLOGY TO BUILD THIS TRACKER
We built this AI safety funding tracker by reviewing every publicly disclosed equity round raised by pure-play or near-pure-play AI safety companies between July 2025 and June 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, 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, token raises, and revenue financing are excluded. Second, we only counted rounds of $300K or more. Third, we only kept companies whose product positioning is primarily AI evaluation, AI red teaming, AI guardrails, AI monitoring, AI risk governance, or model robustness and risk reduction. And fourth, every entry had to be confirmed by a direct company announcement, a press release, or a tier-1 media report, with the source URL preserved for every row.
We excluded generic MLOps, general cybersecurity, general compliance, and general content moderation companies unless the cited product was specifically designed to address AI-model behavior or AI-specific threats. The final dataset contains 20 disclosed deals across 20 unique companies, and every average, median, share, and concentration ratio is computed on that disclosed sample. Privately raised rounds that were never publicly announced are necessarily missing, which is a known limitation of any public-only AI safety funding tracker.
How active has fundraising been in the AI safety market?
As of June 2026, fundraising in the AI safety market has been active but still selective. Over the past 12 months, companies raised 20 disclosed equity rounds and $560.27M combined, which works out to 1.67 deals per month.
The deal count shows real startup formation, but not a flood of fundable companies. The dataset contains 20 unique companies, which means every company appears once rather than being driven by repeat raises during the period.
Dollar flow in the AI safety market averaged $46.69M per month, but that average should not be read as typical. A few large rounds from LMArena, Noma Security, Braintrust, and WitnessAI drive the visible funding signal.
Removing rounds above $50M changes the picture sharply. Without those megarounds, disclosed capital falls from $560.27M to $172.27M, which makes the AI safety market look more like an early-stage formation market than a scaled software category.
If you want to go deeper on the companies shaping this funding activity, see our market report covering AI safety startups.
How concentrated has fundraising been in the AI safety market?
As of June 2026, fundraising in the AI safety market is highly concentrated at the top. Over the past 12 months, the single largest deal accounts for 26.77% of total capital, the top 3 deals reach 58.90%, and the top 5 reach 75.32%.
The largest round is LMArena’s $150M Series A, which alone represents more than one quarter of disclosed capital. That single deal is larger than the entire AI Risk Platforms category and almost twice the AI Guardrail Platforms category.
The top 10 deals account for 91.81% of all disclosed capital. That means the bottom half of the AI safety market, by deal size, accounts for only 8.19% of funding.
This concentration matters because aggregate funding headlines can make the AI safety market look broader than it is. The better reading is that investors have validated a few platform candidates while many companies remain small seed-stage bets.
How much of the AI safety funding signal is driven by outliers?
As of June 2026, a large part of the AI safety funding signal is driven by outliers. Over the past 12 months, only 4 of 20 deals were $50M or larger, but those rounds represented 69.27% of disclosed capital.
The outlier group includes LMArena at $150M, Noma Security at $100M, Braintrust at $80M, and WitnessAI at $58M. Together, these companies shape how the AI safety market looks from a capital perspective.
The median round size is $9.00M, while the average round size is $28.01M. That gap is the clearest warning that the average should not be treated as the typical AI safety financing.
Rounds under $20M still represent most activity by count. There are 14 deals below $20M, which confirms that the AI safety market combines a small number of scaled platform rounds with a much broader layer of early-stage experiments.
This chart, featured in our AI safety market deck, shows how HiddenLayer is positioned in AI safety
Is the AI safety market broad with many targets, or narrow with few fundable companies?
As of June 2026, the AI safety market is narrow rather than broad. Over the past 12 months, the disclosed dataset contains 20 unique companies and 20 deals, which is a meaningful sample but not a deep venture universe.
The narrowness becomes clearer when looking at categories. AI Risk Platforms account for 9 of 20 deals, while AI Guardrail Platforms account for 5. Together, those two categories represent 70.00% of all disclosed activity.
There is more variety in product language than in actual buyer problem. Many companies describe governance, monitoring, guardrails, agent control, and auditability differently, but they often point toward the same enterprise risk-control budget.
The AI safety market is therefore best read as a concentrated formation market. There are enough startups to show founder interest, but not enough scaled companies to treat every segment as validated.
Is AI safety mostly an early-stage formation market or a late-stage scaling market?
As of June 2026, the AI safety market is mostly early-stage by deal count but mixed by capital. Over the past 12 months, Seed rounds represent 10 of 20 deals, while Series A and Series B rounds capture most of the dollars.
Seed rounds account for 50.00% of deal count but only 17.08% of disclosed capital. That means many founders are entering the AI safety market, but most have not yet earned large enterprise-scale rounds.
Series B rounds account for only 3 deals but $238.00M, or 42.48% of total capital. The market has a clear proof premium once companies show they can become default infrastructure.
Early-stage rounds, defined here as Seed, Series A, and Unknown, total $322.27M, or 57.52% of disclosed capital. Late-stage rounds, defined here as Series B and later, total $238.00M, or 42.48%.
For more context on where early AI safety companies are forming, see our deeper analysis of the AI safety market.
Which funding stages are carrying the most capital in the AI safety market?
As of June 2026, Series B and Series A rounds carry most of the capital in the AI safety market. Over the past 12 months, Series B rounds raised $238.00M, while Series A rounds raised $220.40M.
Series B is the largest stage by dollars, with 42.48% of total capital from only 15.00% of deals. That tells us investors are willing to pay up when an AI safety company has stronger enterprise adoption signals.
Series A is almost as important, with 39.34% of total capital from 25.00% of deals. LMArena’s $150M Series A makes this stage unusually large, so Series A should not automatically be read as early validation.
Seed remains the largest stage by count but not by dollars. The median Seed round is $5.75M, which is far below the $80.00M median Series B round and the $18.40M median Series A round.
Which categories attract the most investor attention in AI safety?
As of June 2026, AI Risk Platforms attract the most investor attention by deal count, while AI Evaluation Tools attract the most attention by dollars. Over the past 12 months, AI Risk Platforms closed 9 deals, and AI Evaluation Tools raised $230.00M.
AI Risk Platforms represent 45.00% of disclosed deals but only 21.69% of disclosed capital. That suggests the governance and control-plane layer is crowded with new entrants, but still mostly unproven at scale.
AI Evaluation Tools represent only 10.00% of deals but 41.05% of capital. LMArena and Braintrust show that investors are treating evaluation infrastructure as a potential platform layer, not just a testing workflow.
AI Guardrail Platforms sit in the middle with 5 deals and $78.57M raised. The category has one large Series B in WitnessAI, but most guardrail companies in the dataset remain smaller seed-stage bets.
You can find a deeper category-by-category view in our full market deck on AI safety.
This chart, featured in our AI safety market deck, shows annual funding in AI safety startups
Which categories attract disproportionately large checks in the AI safety market?
As of June 2026, AI Evaluation Tools attract disproportionately large checks in the AI safety market. Over the past 12 months, they represent only 10.00% of deals but 41.05% of disclosed capital.
The capital share to deal share ratio makes the imbalance clear. AI Evaluation Tools score 4.11, while AI Safety Monitoring scores 1.87, and every other category sits below 1.00.
Average deal size also confirms the pattern. AI Evaluation Tools average $115.00M per round, while AI Safety Monitoring averages $52.40M and AI Guardrail Platforms average $15.71M.
AI Red Teaming has strong mindshare but a smaller funding footprint. It represents 10.00% of deals and only 4.53% of capital, which suggests red teaming alone may be treated as a feature unless connected to monitoring, guardrails, or continuous security testing.
Which geographies matter most for fundraising in the AI safety market?
As of June 2026, North America is the geography that matters most for AI safety fundraising. Over the past 12 months, it captured $491.90M, or 87.80% of all disclosed capital.
North America also leads by deal count, with 12 of 20 disclosed rounds. The region’s average round size is $40.99M, and its median round size is $18.20M.
Asia-Pacific contributes meaningful company formation but much smaller rounds. It accounts for 5 deals and $20.37M, or 3.64% of disclosed capital, with an average round size of $4.07M.
Europe has only 2 deals but a healthier capital profile than Asia-Pacific. Geordie AI and White Circle together raised $41.00M, giving Europe 7.32% of disclosed capital and a $20.50M median deal size.
For more detail on the regional funding map, explore our market report covering AI safety geography.
Is the AI safety opportunity set broad globally or concentrated in one hub?
As of June 2026, the AI safety opportunity set is globally visible but heavily concentrated in one hub. Over the past 12 months, North America accounts for 60.00% of deals and 87.80% of disclosed capital.
Asia-Pacific is present but smaller in dollar terms. Its 25.00% deal share shows technical formation in Korea and Australia, but its 3.64% capital share shows much smaller disclosed round sizes.
Europe appears less often but can clear larger checks when the company is tied to agent governance or model control. The region’s two deals average $20.50M, compared with $4.07M in Asia-Pacific.
The Middle East appears through one Israeli seed-stage deal, Willow, at $7M. Latin America and Africa show no disclosed pure-play AI safety equity funding above $300K in this dataset.
This chart, featured in our AI safety market deck, compares the main business model options for AI alignment research labs
Is AI safety a market of small experiments or scaled financings?
As of June 2026, the AI safety market is a market of small experiments plus a few scaled financings. Over the past 12 months, 14 of 20 deals were below $20M, while 4 deals were $50M or larger.
The size distribution is split. There are 5 deals below $5M, 9 deals from $5M to less than $20M, 2 deals from $20M to less than $50M, and 4 deals of $50M or more.
Megarounds over $50M represent 20.00% of deal count but dominate the dollar total. There is only 1 round above $100M, LMArena’s $150M Series A, which represents 5.00% of deals.
The median round size of $9.00M is the best number for understanding the typical company. The average round size of $28.01M is useful for market sizing, but it is not representative of normal fundraising conditions.
If you want to track how this market changes as seed companies graduate, check our AI safety market report.
Who are the investors that appear the most in AI safety fundraising?
As of June 2026, only a few investors appear more than once in AI safety fundraising. Over the past 12 months, Andreessen Horowitz appears in 3 deals, while Insight Partners and Mirae Asset group each appear in 2.
Andreessen Horowitz participated in Promptfoo, LMArena, and Braintrust. That gives it exposure across red teaming, evaluation, and observability, which are three of the most important technical layers in the AI safety market.
Insight Partners appears in Promptfoo and Darwin AI. Its activity spans enterprise AI security testing and government AI governance, which shows interest in both technical assurance and institutional deployment.
Mirae Asset group appears in AIM Intelligence and Tynapse. Both are Asia-Pacific deals linked to AI red teaming, guardrails, or runtime agent security, which makes the pattern more regionally concentrated.
One important caveat is that round announcements rarely disclose individual investor check sizes. Any investor ranking should therefore be read as repeat participation, not as dollars personally committed.
This chart, featured in our AI safety market deck, shows revenue breakdown by customer segment in the AI safety market
INSIGHTS
The insights below come from reviewing every disclosed equity round in the AI safety market between July 2025 and June 2026. They are not row-by-row summaries. They are the reusable patterns that kept showing up across the 20-deal dataset, and they are meant to stay useful when reading any future AI safety funding announcement.
The AI safety market looks much larger in capital than in company count. Four rounds above $50M account for 69.27% of all disclosed dollars. Without those rounds, the market looks like a $172.27M early-stage formation category rather than a fully scaled software market.
AI Evaluation Tools are the clearest platform-scale funding category. The category has only 10.00% of deals but 41.05% of capital. Investors appear to be rewarding evaluation infrastructure as default workflow software, not as a narrow testing service.
AI Risk Platforms dominate formation but not validation. They represent 45.00% of deals and only 21.69% of capital. That gap suggests governance and control-plane startups are abundant, but most are still early in enterprise proof.
The capital share to deal share ratio is the cleanest maturity signal. AI Evaluation Tools score 4.11, while AI Risk Platforms score 0.48 and AI Red Teaming scores 0.45. The ratio separates categories that attract platform checks from categories still funded as exploratory bets.
Series B rounds show the market’s proof premium. They represent only 15.00% of deals but 42.48% of capital. Once an AI safety company proves it can become default infrastructure, investors write much larger checks.
Seed activity shows belief, not maturity. Seed rounds are 50.00% of disclosed deals but only 17.08% of capital. Many founders are entering the AI safety market, but most companies have not yet shown scaled enterprise adoption.
The median round is more informative than the average round. The median is $9.00M, while the average is $28.01M. LMArena, Noma Security, Braintrust, and WitnessAI make the market look bigger than the typical company experience.
The top 10 deals leave very little capital for the rest of the market. They account for 91.81% of disclosed funding. Any market map that treats all funded AI safety startups equally will overstate the breadth of validated demand.
Production-risk adjacency is the strongest buying signal. Companies tied to deployed agents, production AI, observability, runtime controls, and enterprise workflows raised more convincing rounds. Abstract responsible-AI language mattered less than operational control.
AI agents became the market’s central forcing function in 2026. Noma Security, Lumia Security, WitnessAI, JetStream, OpenBox AI, Tynapse, Aigentsphere, Geordie AI, CodeIntegrity, and Willow all frame risk around autonomous or agentic AI. Agent oversight is now one of the market’s clearest demand centers.
The market is moving from pre-deployment assurance toward runtime control. Earlier AI safety narratives focused on evaluation and policy. The 2026 rounds increasingly emphasize monitoring, access control, audit trails, behavior visibility, and intervention at execution time.
There is a sharp distinction between evaluating models and governing deployed systems. LMArena and Braintrust raised large rounds around measurement and observability. JetStream, OpenBox AI, Geordie AI, Willow, and Aigentsphere reflect the separate need to govern agents inside enterprise workflows.
AI Red Teaming has mindshare but limited standalone funding power. The category captured only 4.53% of disclosed capital. Red teaming may be viewed as a feature unless it is paired with continuous testing, monitoring, or guardrail enforcement.
Guardrail platforms show a barbell pattern. WitnessAI raised a $58M Series B, but the median guardrail round is only $5.00M. Most guardrail startups are still experiments, while a small number have broken into enterprise scale.
North America is not just producing more AI safety companies. It is producing the largest validation events. The region captured 87.80% of disclosed capital from 60.00% of deals.
Asia-Pacific shows real formation but smaller capital depth. It has 25.00% of deals but only 3.64% of capital. That suggests meaningful technical activity, especially in Korea and Australia, but smaller disclosed round sizes.
Europe is less frequent but not irrelevant. With only 2 deals, Europe still captured $41.00M. Geordie AI and White Circle show that European AI safety companies can clear larger rounds when tied to agent governance or production control.
Regulatory language is common, but compliance-only stories are not the biggest winners. The largest rounds are tied to operational reliability, security, and control. Buyers appear to fund software that reduces deployment risk, not just policy documentation.
Vertical specialization exists but is not the main capital magnet. ALIGNMT AI focuses on healthcare, and Darwin AI focuses on government. Horizontal platforms for production AI and agents captured larger checks.
The strongest future signal is graduation from Seed to Series A. If 2026 seed-stage agent governance companies raise Series A rounds within 12 to 18 months, the market is converting risk anxiety into recurring software demand. If they do not, 2026 may look more like an exploration wave.
Business Wire (Starseer), Promptfoo (Series A), Noma Security (Series B), ALIGNMT AI (Seed), Wowtale (eRoun&Company), PR Newswire (Darwin AI), Lumia Security (Seed), PR Newswire (LMArena), Axios (WitnessAI), Wowtale (Coxwave), Axios (Braintrust), JetStream Security (Seed), OpenBox AI (Seed), Seoul Economic Daily (AIM Intelligence), Wowtale (Tynapse), Startup Daily (Aigentsphere), Business Wire (White Circle), GeekWire (CodeIntegrity), Geordie AI (Series A), PR Newswire (Willow)
Related blog posts
- All funding deals 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?
