Barcelona startup Galtea just locked down $3.2 million in seed funding to tackle one of the most expensive and overlooked problems in artificial intelligence: testing AI agents before they go live. With over 80% of AI projects failing to deliver real business value, the timing could not be sharper.
Here is what this deal means for the future of enterprise AI and why it matters right now.
What Galtea Does and Why It Matters
1 Galtea helps enterprises stress-test AI agents before deployment, aiming to cut the risks that keep an estimated 95% of AI projects from reaching production. 11 According to the company, a major bottleneck in the AI industry is the availability of accessible and trustworthy test data.
Think about it this way. 1No one tests a new self-driving algorithm directly on the road. Instead, companies log thousands of hours in simulations with synthetic environments that reveal failures real-world tests would take years and accidents to find. Galtea applies that same philosophy to AI agents used in banking, healthcare, telecom and education.
2 The platform generates test cases and synthetic user simulations from descriptions of how an AI agent is intended to behave, creating adversarial and edge-case scenarios automatically, at scale, without requiring engineering teams to write them by hand.
11AI testing alone costs an estimated $13 billion annually for companies across Europe and the US. That is a staggering number, and it underscores exactly why investors are paying attention.
Galtea AI testing platform seed funding round 2026
Who Backed the Round and What the Money Will Fund
1 The $3.2 million seed round was led by German fund 42CAP, with participation from Mozilla Ventures and existing investors JME Ventures, Masia and ABAC Nest Ventures. 1 This brings Galtea’s total funding to $4.1 million, following an $870,000 pre-seed round in 2024. 2 The Mozilla Ventures involvement signals that the round was framed partly around the trustworthy-AI narrative that has defined the fund’s portfolio thesis since its launch in 2022.
Julian von Fischer of lead investor 42CAP put it plainly: 4“As enterprises race to deploy generative AI, the gap between what models can do and what companies can trust them to do is widening fast. Jorge and Baybars have built the missing quality assurance layer.”
Here is how Galtea plans to use the fresh capital:
- Engineering expansion to accelerate platform development
- Commercial team growth to build a stronger sales engine
- Geographic expansion with a focus on the UK market alongside its base in Spain and Portugal
- Self-service product launch including a two-week free trial for developers
4 The company’s workforce now stands at 12 after doubling over the past year.
Real Results From Real Customers
Galtea is not just pitching a concept. 1Early enterprise customers, including Telefonica and Spanish fintech ABANCA, are already using the platform in production and pilot deployments.
The numbers are hard to ignore.
| Metric | Result |
|---|---|
| Average cost reduction in AI validation | 71% |
| Estimated return on investment | 10X |
| Test scenarios auto-generated (one case study) | 6,000+ |
| Manual hours saved (one case study) | ~600 hours |
| Failed evaluations found vs. internal testing | 12X more |
6 In one case study, a customer support agent tested via Galtea failed over 2,100 evaluations related to critical vulnerabilities, a failure rate 12 times higher than internal testing indicated. 6 The platform automatically generated over 6,000 test scenarios for this agent, saving approximately 600 hours of manual effort and preventing potential production issues.
Jorge Romaris, Head of AI at ABANCA, confirmed the impact: 4“With Galtea, we uncovered vulnerabilities we would likely have missed otherwise, saved significant engineering time, and improved the reliability of our AI systems.”
The Two Failure Modes Every AI Team Should Know
CEO Jorge Palomar, who previously worked at Amazon, and CTO Baybars Kulebi, a PhD holder in Astrophysics, co-founded Galtea after meeting at the Barcelona Supercomputing Centre. 2The technology was originally developed at BSC to evaluate large language models for internal research purposes, running on MareNostrum 5, one of Europe’s most powerful supercomputers.
Palomar says two failure modes come up again and again in customer deployments.
First: hallucination with proprietary data. 1“If you ask about the refund policy and it doesn’t find anything, it will make up a ‘reasonable’ refund policy that may not be the one for that company. That’s a very, very common error,” Palomar explained.
Second: regulatory non-compliance. 1Industries like finance, healthcare and education have strict rules against certain advice or recommendations. The LLM does not know these rules and might generate financial advice, medical recommendations or legal opinions. The deployment system should block this, but testing often misses some cases.
Both of these failures can lead to lawsuits, fines and a total loss of customer trust. And both are preventable with proper testing before launch.
Why EU AI Act Pressure Makes This Urgent
Galtea’s timing is not accidental. 2The EU AI Act now requires companies deploying AI in high-risk applications to document and validate their models’ safety and compliance, with fines of up to 35 million euros for violations.
29The most impactful deadline for enterprises is August 2, 2026, when high-risk AI system requirements become enforceable. That is just months away.
29 Organizations deploying AI in healthcare, finance, law enforcement, education or critical infrastructure must be compliant by this date. 28 Analysis of organizational readiness suggests most enterprises face significant compliance gaps as the deadline approaches.
For context, consider the broader AI landscape right now:
- 19 95% of GenAI pilots fail to scale, and over 80% of all AI projects fail overall
- 19 In 2025, global enterprises invested $684 billion in AI initiatives. By year-end, over $547 billion of that had failed to deliver intended business value
- 19 The average sunk cost per abandoned AI project sits at $4.2 million
2 For European enterprises that have been building AI products without systematic testing infrastructure, the regulation has created urgency. Galtea’s platform sits directly in that gap, helping development and legal teams produce the evidence of compliance the regulation demands.
“Most teams ship AI agents with a handful of hand-written test cases and hope nothing breaks in production.” — Galtea website
1 If successful, Galtea’s approach could shift AI development from reactive monitoring to proactive validation. Right now, most AI teams see testing as a tax, something done reluctantly and minimally at the end. Palomar argues it should be the foundation you build around.
As AI agents become more powerful and more autonomous, the question is no longer whether companies can build them. It is whether they can prove those systems work safely before putting them in front of real customers, real patients and real money. Galtea is betting $4.1 million that the answer starts with better testing. And looking at the numbers, it is hard to argue otherwise. What do you think? Drop your thoughts in the comments below.
