What are the current funding trends in AI safety?

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.

AI safety funding: what is actually changing in 2026?

Is AI safety funding really booming right now?

AI safety funding is growing fast in 2026, and this time the boom goes well beyond a few giant rounds.

We counted 56 financings in the current year-to-date period, up from 42 over the same months of 2025 and 33 in 2024. The number of funded companies also rose from 42 to 52 in one year. At the same time, disclosed capital reached $1.44 billion, compared with $571 million a year earlier.

The mix is what makes 2026 unusual. More startups are raising, but the companies that do raise are also getting much bigger checks. The median financing climbed from $5.68 million to $10 million, while the average moved from $14.3 million to $25.7 million.

The 2024 comparison is useful because that year included Safe Superintelligence's $1 billion financing. Despite that enormous outlier, 2026 has already produced almost 70% more deals over the comparable period and 62.5% more funded companies.

Market measure	2026 YTD	2025 YTD	2024 YTD	Change vs 2025
Deals	56	42	33	+33.3%
Unique funded companies	52	42	32	+23.8%
Total disclosed capital	$1.440B	$571M	$1.323B	+152.1%
Capital excluding >$50M rounds	$625M	$221M	$263M	+182.5%
Capital excluding >$100M rounds	$1.062B	$571M	$323M	+85.9%
Median round	$10.0M	$5.68M	$7.25M	+76.2%
Average round	$25.7M	$14.3M	$41.3M	+80.1%
First-financing share	48.2%	57.1%	57.6%	-8.9 pp
Follow-on share	51.8%	38.1%	42.4%	+13.7 pp

Would AI safety funding still look strong without the biggest deals?

Yes. AI safety funding actually looks even stronger once we strip out the largest rounds.

Removing every financing above $50 million leaves $625 million of capital in 2026. The comparable figure last year was $221 million, so funding below that threshold has almost tripled.

The same test helps with 2024. Headline capital from that period looks close to today's level because Safe Superintelligence raised $1 billion by itself. Below $50 million, current funding is about 138% higher than the comparable 2024 figure.

Money has also spread much further down the ranking. The five largest deals currently account for 38.8% of capital, versus 65% in 2025 and 85.3% in 2024. Goodfire's $150 million Series B, Zenity's approximately $115 million of primary capital, Onyx Security's $113 million Series B, HiddenLayer's $100 million Series B and Braintrust's $80 million Series B are all large, but together they still leave most of this year's funding elsewhere in the market.

The market is simply much less top-heavy than it was in 2025 or 2024.

Are normal AI safety rounds getting much bigger?

Yes. The clearest change in AI safety funding right now is happening in the $20 million to $50 million range.

We found 12 financings in that bracket this year. There was only one during the comparable 2025 period. Six more deals landed between $50 million and $100 million, compared with two last year.

Meanwhile, sub-$5 million rounds are becoming less dominant. They represented 42.9% of 2025 deals and now account for 25%.

There are now far more companies raising $20 million, $30 million, $40 million or more. A couple of years ago, the market had a much larger population of small early rounds and an occasional spectacular financing at the top.

Twenty-two of this year's 56 deals are worth at least $20 million. Only five of 42 crossed that mark in the same part of 2025.

Round size	2026 deals	2026 deal share	2026 capital share	2025 deals
Below $5M	14	25.0%	2.4%	18
$5M to <$20M	20	35.7%	12.4%	17
$20M to <$50M	12	21.4%	25.1%	1
$50M to <$100M	6	10.7%	26.9%	2
$100M+	4	7.1%	33.2%	2
At least $20M	22	39.3%	85.2%	5
Below $20M	34	60.7%	14.8%	35
Median round	$10.0M	—	—	$5.68M

Is AI safety finally moving beyond Seed?

Yes. AI safety is still full of young startups, but almost all of the extra deal activity this year has appeared after Seed.

Seed financings stayed exactly flat at 36. Total deals still rose from 42 to 56 because Series A jumped from two financings to 12, Series B from one to five and Series C from one to two.

Series A has become especially important. Those 12 deals brought in about $425 million, with a $30 million median round. Five Series B financings raised $493 million between them, more than all 36 Seed deals combined.

The speed is striking. The typical Series A company in our research is only around two years old, while the typical Series B company is roughly three. Startups founded during the recent generative-AI wave are reaching serious institutional funding rounds very quickly.

Seed still represents almost two-thirds of transactions, so startup formation remains active. The big change is that there is now a much larger class of young companies successfully graduating beyond it.

Can brand-new AI safety startups still get funded?

Yes, new AI safety startups are still getting funded, although most of the extra money is now going to companies investors have backed before.

We identified 27 first financings this year, compared with 24 during the same part of 2025. New-company funding has therefore grown slightly in absolute terms.

Follow-ons tell a much more dramatic story. Their count rose from 16 to 29, and the median follow-on increased from $7 million to $30 million.

Capital followed the same path. First financings brought in around $232 million, barely above the $222 million recorded a year earlier. Follow-ons climbed from $262 million to roughly $1.21 billion.

Investors have not stopped funding new teams. They are putting far more money behind companies that have already passed an earlier fundraising and product-development test.

Our first-financing classification comes from a financing-history check rather than a claim that we can reconstruct every private round ever completed by every company. The gap is large enough, however, that the direction is clear.

Financing profile	2026	2025	2024
First financings	27	24	19
First-financing share	48.2%	57.1%	57.6%
First-financing capital	$232M	$222M	$1.139B
Median first financing	$6M	$4M	$7M
Follow-on financings	29	16	14
Follow-on share	51.8%	38.1%	42.4%
Follow-on capital	$1.208B	$262M	$184M
Median follow-on	$30M	$7M	$10M
Companies with multiple deals in-period	4	0	1

Why are AI guardrail startups getting so much money now?

AI guardrails are currently the biggest funding story inside AI safety, with investors backing many more companies and writing much larger checks.

We counted 22 Guardrail Platform financings this year, compared with seven in the same part of 2025. Their share of all deals rose by 22.6 percentage points, while their share of capital increased by 32.7 points.

The median guardrail round has also moved up to $19.5 million from $9 million.

This change has been building for a while. Lakera raised $20 million in 2024 to protect generative-AI applications from prompt injection, data leakage and other AI-specific attacks. Prompt Security also raised Seed and Series A capital around enterprise GenAI controls. Those companies were early signs of the category.

The 2026 market is much bigger. Zenity's latest financing included roughly $115 million of primary capital for a platform designed to discover enterprise agents and put security boundaries around what they can do. Onyx Security raised $113 million after previously raising a $35 million Series A. The company's platform discovers AI agents, watches their behavior and enforces runtime controls.

This is broad category growth rather than one financing distorting the totals.

AI safety category	2026 deals	2026 capital	Median round	Deal-share change vs 2025
AI Guardrail Platforms	22	$626M	$19.5M	+22.6 pp
AI Risk Platforms	11	$212M	$5.8M	-6.5 pp
AI Safety Monitoring	9	$172M	$8.0M	+4.2 pp
AI Red Teaming	6	$89.6M	$8.5M	-3.6 pp
AI Evaluation Tools	5	$171M	$28M	-12.5 pp
Model Robustness Tools	3	$170M	$18.5M	-4.2 pp
All AI safety deals	56	$1.440B	$10.0M	—

Is runtime AI security taking over from evaluation and red teaming?

Runtime AI security is taking a much larger share of funding today, while evaluation and red teaming are evolving in a more uneven way.

Safety Monitoring went from five deals to nine and from roughly $19 million to $172 million of capital. HiddenLayer is a useful example. Its recent $100 million Series B is specifically being used to deepen agentic runtime protection, and the company reported more than tenfold ARR growth plus over 50 new platform customers.

AI Evaluation looks different. We counted five deals this year versus nine last year, but its median financing jumped from $3.3 million to $28 million. Braintrust's $80 million Series B fits that pattern. The company says AI has moved from experimentation into production, where long-running agents now generate complex traces and failure modes that teams need to monitor continuously.

Red Teaming has not vanished either. The number of deals stayed at six, although its share of the expanding overall market declined.

More money is now going toward what happens after an AI system goes live. Testing before deployment still matters, but production observability, agent control and runtime protection are absorbing a larger portion of venture capital.

Is model interpretability having the same funding boom as guardrails?

Model interpretability is attracting serious money, but the current funding spike is concentrated in one standout company.

Model Robustness Tools generated roughly $170 million from only three financings. Goodfire alone raised $150 million in its Series B after previously raising a $50 million Series A.

Goodfire is building interpretability technology designed to understand and shape what happens inside advanced models. Its latest funding will support research as well as applications involving AI agents and life sciences.

Three financings do not give us the same breadth we see in guardrails or monitoring. The category's capital share has risen sharply because Goodfire became a nine-figure company very quickly.

For now, model interpretability looks like a high-conviction pocket of the market rather than evidence that dozens of robustness startups are suddenly breaking out.

Why is money moving toward AI agent security?

Money is moving toward AI agent security because agents can now touch real company data and systems, creating a much harder security problem than a chatbot answering questions.

Microsoft's February 2026 Cyber Pulse research said more than 80% of Fortune 500 companies were deploying active agents built with low-code or no-code tools. Microsoft also warned that agent adoption was moving faster than some companies' ability to observe and govern it.

Google Cloud reached a similar conclusion in its 2026 AI infrastructure research. Among the technology leaders it surveyed, 79% cited security, governance or operations as their biggest challenge to scaling inference. Google also reported that 35% of senior IT decision-makers saw inadequate security for multi-system access as a major obstacle to deploying agents.

An agent may be allowed to read email, query a database, use a tool, call an API or trigger an action. Every additional permission creates something that needs to be identified, monitored and controlled.

Funding is following that problem. Onyx is building an AI control plane, HiddenLayer is pushing further into runtime agent protection, and Zenity is focused on agent discovery, governance and enforcement.

These products are easier to justify when thousands of autonomous or semi-autonomous agents are entering corporate systems than when generative AI mostly lived inside isolated pilots.

Are AI safety startups getting older as the market matures?

No. AI safety startups are reaching bigger rounds remarkably quickly, while the typical funded company remains very young.

The approximate median company age in this year's financing cohort is only one year. It was two years in 2025 and one year in 2024.

That sits oddly beside the stage shift at first. Series A and Series B are much more common, yet the companies raising them are not old cybersecurity vendors that have suddenly repositioned around AI.

The typical Seed company is around one year old, Series A around two and Series B around three. Even within Guardrail Platforms, the median company age is roughly one year.

The market is maturing financially faster than the startups themselves are aging. A recent cohort is moving through the fundraising cycle unusually fast.

Is North America taking over AI safety funding?

North America currently dominates AI safety capital, although Europe is producing more funded companies than it did last year.

North America generated 34 deals, giving it 60.7% of total activity. That is actually lower than its 66.7% deal share in 2025. Its capital share moved the other way, from 58.8% to 83.1%.

The median North American round reached $22.5 million, up from around $5 million a year earlier.

Europe doubled from six deals to 12, pushing its deal share to 21.4%. The median European financing remained around $4.5 million, so the region's growth currently comes more from having additional funded startups than from producing many huge rounds.

The Middle East contributed seven deals and $132 million, while Asia-Pacific remained smaller with three deals and about $13 million.

We also found roughly 10 countries represented this year compared with around seven in 2025. AI safety is spreading geographically even as the largest checks cluster more heavily in North America.

Are more venture investors seriously betting on AI safety now?

Yes. More investors are leading AI safety rounds today, and several firms are coming back for multiple deals.

We found 55 unique disclosed lead investors in this year's sample, up from 46 last year. Six leads appeared on more than one financing, compared with only one repeat lead in the comparable 2025 period.

Andreessen Horowitz appeared three times as a lead or co-lead. SYN Ventures, Hetz Ventures, Norwest, Bessemer Venture Partners and Ugly Duckling Ventures each appeared twice.

New investors are still entering the sector, while a group of existing backers is gaining enough conviction to fund several companies.

A year ago, Race Capital was the only lead that appeared repeatedly in our comparable sample. Its absence from this year's lead list tells us only that it did not lead one of the deals we captured during the period; it would be too much to call that a retreat from AI safety.

Overall, venture interest looks broader and more repeatable than it did a year ago.

Why are corporate investors showing up in more AI safety deals?

Corporate investors are much more visible in AI safety rounds this year, especially once startups reach Series A and beyond.

Strategic investors participated in 16 financings, or 28.6% of all deals, compared with five deals and 11.9% last year.

The more interesting change is where those investments sit. Only four of the 16 strategic-backed rounds were Seed deals. Six were Series A, three Series B and two Series C.

The median strategically backed financing reached $37 million.

Microsoft's M12 and Booz Allen Ventures joined HiddenLayer's $100 million Series B. Zenity's latest round included Hitachi Ventures, LG Technology Ventures and Intel Capital. Goodfire's investor group included Salesforce Ventures. These companies sit close to cloud infrastructure, enterprise software, security or AI deployment, so the strategic link is straightforward.

Corporate participation was already high in 2024 at about 30% of deals, which keeps us from calling the current level unprecedented. What looks different now is the amount of money involved and the maturity of the rounds.

Is regulation actually causing the AI safety funding boom?

Regulation is helping create demand for AI governance and safety tools, but commercial deployment of AI agents looks like the more immediate force behind this year's funding shift.

The regulatory pressure is real. General-purpose AI obligations under the European AI Act began applying in 2025, including requirements around technical documentation, transparency, risk management and safety for the most advanced models.

The framework became more concrete in 2026. New transparency requirements started applying, and the European Commission's enforcement powers expanded under the AI Act.

That gives companies another reason to document, monitor and control their AI systems.

Still, the category changes in our funding research line up more directly with production problems. Guardrails and Safety Monitoring are gaining deals. HiddenLayer talks about securing autonomous coding agents at runtime. Onyx is selling control over enterprise agents. Microsoft's and Google Cloud's latest research both focus heavily on agent permissions, observability and governance.

Regulation adds pressure, especially in Europe. The current financing pattern looks broader than compliance spending alone.

Has AI safety funding already passed previous full-year levels?

AI safety funding has already beaten some full-year benchmarks, even before 2026 is complete.

Current year-to-date capital equals roughly 175% of all the funding we recorded during 2025. The 56 deals completed so far are already about 90% of the 62 financings recorded during that entire year.

Activity has also moved past 2024. We counted 52 deals across all of 2024, so the current year has already exceeded that total.

Capital has not yet passed full-year 2024 because Safe Superintelligence's $1 billion round made that year unusually large. Even so, current funding has already reached roughly 91% of the 2024 total.

We would not extend these numbers into a mechanical year-end forecast. Venture funding arrives in bursts, and a couple of large financings can move a yearly total dramatically. What we can say is that the current pace has already crossed benchmarks that previously required a full calendar year.

What is actually changing in AI safety funding?

The biggest change is the arrival of a real scale-up layer. AI safety used to lean much more heavily toward Seed rounds and a few huge outliers. Today, many more companies are raising $20 million to $50 million, Series A activity has surged and Series B is becoming much more visible.

The second change is where the money goes after the first round. New startups continue to enter the market, but follow-on capital has accelerated far faster. Investors are now putting much larger amounts behind companies that have already shown product or commercial progress.

The clearest category rotation is toward guardrails, monitoring and runtime control. These products are closely tied to the move from experimental generative AI toward agents operating inside real enterprise systems. Evaluation remains well funded at the upper end, while model interpretability has produced a few high-conviction bets rather than the same broad expansion.

The outlier test is unusually reassuring here. Capital excluding rounds above $50 million is growing even faster than headline capital, while top-five concentration has fallen sharply. The boom survives once we remove the deals most likely to distort it.

The part we would watch most carefully is startup formation. Seed activity itself has stopped growing, even as later stages accelerate. If first financings and Seed stay flat for much longer, the market could eventually become more dependent on a cohort that was created during the 2023–2025 wave.

For now, AI safety funding is moving from startup formation into fast scaling, with AI agent control becoming the part of the market attracting the broadest investor attention.