AI Safety Funding: A Quarter-by-Quarter Analysis

All funding deals in the AI safety market over the last years

Below is the table listing all the deals. You can find our methodology at the end of this page.

If you want a deeper understanding of the market and its current dynamics, get our report covering the AI Safety Market.

Is AI safety funding actually getting stronger right now?

AI safety funding is stronger now than the headline quarterly totals make it look.

Across the five complete quarters we studied, funding moved from $203.3M in Q2 2025 to $253.0M in Q3, $243.0M in Q4, a record $464.1M in Q1 2026, then back to $373.9M in Q2. Looking only at that sequence makes the market seem erratic.

The underlying financing environment has improved much more clearly. Once we exclude rounds above $50M, funding reached $249.9M in Q2 2026, more than double Q2 2025 and the highest level in the period. The median round also reached $10.5M.

Q1 2026 produced the biggest headline number, but Q2 had more depth. Money was spread across a wider group of companies raising substantial institutional rounds instead of depending on one $100M-plus financing.

Period	Deals	Capital	Capital excluding >$50M	Median round
Q2 2025	14	$203.3M	$103.3M	$6.55M
Q3 2025	18	$253.0M	$73.0M	$3.53M
Q4 2025	18	$243.0M	$162.0M	$8.50M
Q1 2026	19	$464.1M	$176.1M	$6.00M
Q2 2026	18	$373.9M	$249.9M	$10.50M
Q2 2025 → Q2 2026	+28.6%	+83.9%	+141.9%	+60.3%
Q1 → Q2 2026	−5.3%	−19.4%	+41.9%	+75.0%

When did AI safety funding really start to change?

The AI safety funding market really began to change in Q4 2025.

Q3 initially looked like an acceleration because total funding rose 24%. Yet ordinary financing was getting weaker: the median round fell 46% to $3.53M and capital excluding rounds above $50M dropped 29%.

Q4 flipped that pattern. Total funding barely moved, from $253.0M to $243.0M, while the median more than doubled to $8.5M. Capital outside the very largest rounds jumped from $73.0M to $162.0M.

We also saw a change in what companies were raising money for. Guardrail platforms went from one financing in Q3 to nine in Q4, while follow-on rounds became much more prominent.

Q4 was more important than the raw funding chart suggested. It was the first quarter where bigger checks started appearing across the market rather than mainly at the top.

Was Q3 2025 basically just two huge AI safety deals?

Yes, much of Q3 2025's apparent AI safety funding strength came from two unusually large rounds.

Noma Security raised $100M and Irregular raised $80M. Together they represented roughly 71% of all capital raised that quarter.

The rest of the market looked very different. Twelve of the 18 financings were below $5M, and the median round fell from $6.55M in Q2 to only $3.53M.

Q3 therefore combined two things that can easily be confused: more companies were getting funded, but most were raising small rounds. The two large transactions then pushed the headline total above Q2.

The same effect distorted the geographic picture. Because both Noma Security and Irregular were classified in the Middle East, the region suddenly accounted for 71% of quarterly capital. That concentration disappeared in the following quarter.

Q3 had plenty of activity, but broad-based funding had not taken off yet.

Did Q1 2026 really mark an AI safety funding boom?

Q1 2026 was a big scaling quarter for a few AI safety companies, while the broader market grew much more modestly.

Goodfire raised $150M, Braintrust raised $80M and WitnessAI raised $58M. Those three financings alone generated about 62% of the quarter's $464.1M total.

Deal count barely changed, moving from 18 to 19. The average financing jumped to $24.4M, yet the median fell to $6M.

A handful of companies had reached the point where investors were willing to write very large growth checks, while the typical deal remained far smaller.

Later-stage funding also reached 56% of quarterly capital, largely because of those larger rounds.

Q1 was important because it showed that some AI safety companies could already support $50M-to-$150M financings. Calling it a market-wide boom would go too far.

Why did Q2 2026 look weaker when the AI safety market was actually healthier?

Q2 2026 was healthier than its 19% drop in headline AI safety funding suggests.

The largest financing in Q2 was $64M and represented only 17% of quarterly capital. In the four earlier quarters, the biggest deal alone accounted for between 32% and 49%.

That gave Q2 a much broader funding base. Seven transactions reached at least $20M, covering companies in guardrails, red teaming and AI evaluation.

At the same time, the market no longer needed a $100M or $150M transaction to produce a large quarterly total.

Q2 currently looks like the strongest underlying quarter in the five-quarter period. The headline number fell because Q1 had several exceptional financings, while the rest of the market kept moving up.

Are AI safety investors writing much bigger checks now?

Yes, ordinary AI safety funding rounds got substantially bigger by Q2 2026.

Only two transactions were below $5M in Q2, down from eight in Q1 and 12 back in Q3 2025. Nine companies raised between $5M and $20M, four raised between $20M and $50M, and three landed in the $50M-to-$100M range.

That distribution is much more useful than a single blockbuster deal. Companies were increasingly able to raise checks in the tens of millions across several parts of the market.

The $20M-to-$50M bucket is especially telling. It represented almost one-third of Q2 capital after being absent altogether in Q3 2025.

The latest quarter also had no $100M-plus financing. Bigger ordinary rounds, rather than one huge outlier, carried the market.

Round size	Q2 2025	Q3 2025	Q4 2025	Q1 2026	Q2 2026
<$5M	6 deals	12	7	8	2
$5M-<$20M	6	4	8	5	9
$20M-<$50M	1	0	2	3	4
$50M-<$100M	0	1	1	2	3
$100M+	1	1	0	1	0
Deals of $20M+	2	2	3	6	7
Total deals	14	18	18	19	18

Is Series A becoming the main battleground in AI safety funding?

Series A is becoming the key funding stage in AI safety, and companies are reaching those large rounds while they are still remarkably young.

Six Q2 2026 transactions were Series A financings. Together they raised $229M, or 61% of all capital in the quarter, with a $35M median round.

Series A represented only 7% of deals in Q2 2025. By the latest complete quarter it represented one-third.

Yet company age has barely moved. The median company receiving funding has been one year old since Q3 2025. Follow-on companies are generally only two or three years old.

Gray Swan, founded in 2023, raised a $40M Series A. Coval, founded in 2024, raised $28M. Even among younger businesses, investors are increasingly comfortable writing institutional-size checks once commercial or technical traction is visible.

Maturity in AI safety is happening very quickly. Companies do not need to become old businesses before they can raise serious money.

Are first-time AI safety startups getting squeezed out?

No, first-time AI safety startups are still getting funded, but they receive a much smaller share of the money than companies coming back for another round.

Seven of the 18 Q2 2026 transactions were first financings. That is still a meaningful pipeline of new companies.

Capital has moved much more decisively. Follow-on financings accounted for 32% of funding in Q2 2025, then rose to 52% in Q3, 77% in Q4 and just over 81% in both 2026 quarters.

By Q2 2026, first financings represented 39% of deals but only 19% of capital.

We still see plenty of startup formation. What has changed is where investors put the larger checks: increasingly into companies that have already raised once, built a product and returned with more evidence that customers actually want it.

That distinction is more useful than saying investors have simply become less interested in early-stage AI safety.

Are AI guardrail platforms taking over AI safety funding?

AI guardrail platforms currently have the strongest momentum in AI safety funding, with enough separate deals to make the move harder to dismiss as one-company noise.

Guardrails accounted for 10 of 18 Q2 2026 financings and $193.3M of capital. Straiker and Arcade.dev provided two of the largest rounds, while several smaller companies added depth below them.

The category has been volatile. Guardrails represented only one deal in Q3 2025, jumped to nine in Q4, pulled back in Q1 and then reached 10 in Q2.

Still, the latest quarter looks stronger than a typical category spike because several companies raised meaningful amounts at the same time.

Model robustness gives us a useful contrast. Goodfire's $150M round made that category look huge in Q1, then model robustness recorded no Q2 financing at all. AI evaluation has also produced sharp jumps around individual companies such as LMArena, Braintrust and Patronus AI.

Guardrails currently have both deal breadth and capital depth. We need another few quarters before treating that dominance as permanent.

Q2 2026 category	Deals	Capital	What we see now
AI Guardrail Platforms	10	$193.3M	Broadest momentum
AI Evaluation Tools	2	$78.0M	Large rounds, fewer companies
AI Red Teaming	3	$57.0M	Re-accelerating
AI Risk Platforms	2	$32.6M	Clear recent slowdown
AI Safety Monitoring	1	$13.0M	Still small and uneven
Model Robustness Tools	0	$0	Q1 surge did not persist
Total market	18	$373.9M	—

Why is agent security suddenly pulling in so much AI safety money?

Agent security is pulling in capital because enterprise AI is moving from systems that answer questions toward systems that can take actions inside real software.

That changes the security problem. Companies now have to think about which tools an agent can call, what data it can access, what permissions it receives and what happens when an autonomous workflow behaves unexpectedly.

The timing lines up closely with the funding shift. The NSA released dedicated Model Context Protocol security guidance in May 2026, pointing to accelerating real-world adoption and risks around tool invocation, trust relationships and context sharing. OWASP has also built a dedicated Agentic Security Initiative around autonomous agents and multi-step workflows.

The companies raising money were working directly on those problems. Arcade.dev's $60M financing centered on authorization and control of actions taken by agents. Straiker's $64M Series A funded a platform combining discovery, adversarial testing and runtime protection, after the company said run-rate revenue had increased more than 15-fold in less than a year.

WitnessAI had already expanded its product toward agent governance earlier in 2026, including visibility into the tools and systems enterprise agents can access.

We cannot prove that agent adoption caused every funding decision. But the match between the new technical problem and the companies receiving large rounds is unusually strong.

Are corporate investors becoming a serious force in AI safety funding?

Corporate investors are now a serious part of AI safety funding, especially once startups reach the stage where their products are touching real enterprise infrastructure.

Strategic investors appeared in about 6% to 11% of deals during 2025. That share jumped to 26% in Q1 2026 and 39% in Q2.

Seven Q2 financings included an identifiable strategic investor, and five of those seven were Series A rounds. The median strategic-backed round reached $40M.

The investor mix also makes commercial sense. Samsung Venture Investment backed AIM Intelligence. Datadog and Samsung participated in Patronus AI. Earlier in the year, Qualcomm Ventures and Samsung Ventures joined WitnessAI's strategic round.

These investors sell hardware, enterprise software, observability tools, communications infrastructure or other products sitting close to AI deployment.

Arcade.dev also brought in Morgan Stanley and Wipro, while Straiker attracted Citi Ventures and Workday Ventures. The common thread is practical: companies with direct exposure to enterprise AI are putting money into the controls needed to deploy it safely.

Strategic participation is one of the clearest structural changes we found.

Are the same VCs starting to control AI safety funding?

No, the same small group of VCs is not taking over AI safety funding; lead investors are still spread across a surprisingly wide set of firms.

We identified 14 unique disclosed leads in Q2 2025, followed by 20 in Q3, 26 in Q4, 18 in Q1 2026 and 22 in Q2.

Repeated leadership within a single quarter is unusual. Race Capital appeared on three Q2 2025 rounds, while Q3, Q4 and Q1 2026 had no comparable cluster. SYN Ventures led two transactions in Q2 2026.

A few investors do appear across several quarters. Redpoint Ventures shows up in three consecutive quarters, while Insight Partners and SYN Ventures each recur across two.

The broader picture still looks fragmented. AI safety companies are raising larger rounds without becoming dependent on a tiny circle of specialist investors.

Capital is concentrating around companies that have already proved something, while the firms supplying that capital remain fairly diverse.

Is AI safety funding becoming less US-centric?

AI safety funding is becoming more geographically spread by deal count, although North America still dominates the money.

North American companies represented 64% of deals in Q2 2025 and 50% in Q2 2026. That is a real change in participation.

Capital is moving more slowly. North America still captured 73% of Q2 2026 funding.

Europe has the clearest momentum outside North America. After a small presence in Q2 2025, European companies accounted for roughly 22% to 28% of deals in every later quarter. By Q2 2026, Europe was also approaching one-fifth of total capital.

France, Spain and the UK all produced financings during the latest quarter, including an $11M round for White Circle.

Asia-Pacific remains much smaller. The Middle East can look large in isolated quarters, but the earlier 71% Q3 capital share came almost entirely from two deals and disappeared immediately afterward.

The market is becoming more international, especially on deal count. North America's funding lead remains comfortable for now.

Which AI safety funding trends are actually real today?

The two clearest AI safety funding trends today are the rise of follow-on capital and the much bigger role of strategic investors.

Both changes develop across several quarters instead of appearing once and disappearing.

Larger ordinary rounds also look increasingly convincing. We first saw that shift in Q4 2025, and Q2 2026 provided much stronger confirmation after the megaround-heavy Q1.

Series A is following the same path. The latest quarter was unusually strong, but we would still want more history before calling the shift permanent.

Guardrail platforms currently have the best category momentum, while Europe's growing capital depth is another development worth watching.

Several dramatic-looking patterns do not survive the same test. The Middle East surge was concentrated in one quarter. Model robustness spiked around Goodfire. Q1's jump in later-stage funding reversed quickly. Lead-investor concentration never really appeared.

Apparent trend	Our current read
More capital going to follow-on rounds	Persistent
More strategic-investor participation	Persistent
Bigger ordinary round sizes	Emerging, increasingly convincing
Series A taking more capital	Emerging
Guardrails leading the category mix	Strong current momentum, still volatile
Europe attracting deeper funding	Emerging
First financings disappearing	Not supported
Q1 2026 later-stage surge	Temporary spike
Q3 2025 Middle East surge	Temporary spike
Model-robustness surge	Temporary spike
Lead investors becoming concentrated	Not supported

What is the funding trend actually telling us?

AI safety funding has moved from a market where quarterly totals were easily distorted by one or two huge rounds toward one where a broader group of companies can raise checks in the tens of millions. The path has been messy, but the financing base is stronger than it was at the start of the period.

Capital increasingly follows companies that have already raised once. New startups continue to appear, while repeat fundraisers capture most of the dollars because they can show more product, customer or technical evidence.

Series A has become especially important. Young AI safety companies are reaching large institutional rounds remarkably quickly, so the market is maturing through financing progression rather than through an aging company base.

Guardrails and agent security currently have the strongest category momentum. That funding wave also fits what is happening outside venture capital: autonomous agents introduce new problems around permissions, tool use, runtime behavior and control, which enterprises increasingly need to solve before putting these systems into production.

Corporate money is becoming harder to ignore. Banks, software companies, communications providers and technology groups are investing in the same security layer they may eventually depend on as AI agents spread through enterprise systems.

The raw quarterly funding chart remains the easiest part of this market to misread. Q3 2025 looked healthier than it was, while Q2 2026 looked weaker than it was. Once we separate exceptional rounds from the rest of the market, the direction becomes much clearer.