Tangos AI Inc., a startup that uses artificial intelligence to conduct financial crime investigations at unprecedented scale, said today it has closed on $20 million in seed funding.

The round was led by Red Dot Capital Partners and saw participation from Leaders Fund, Clarim, Venture Israel, Signal Fire, Clutch Capital and Selah Ventures, as well as Bright Data Inc., which made a strategic investment into the company.

Tangos believes there’s a big opportunity to help financial institutions, government agencies and intelligence organizations crack down on financial crime. The United Nations estimates that the amount of money laundered globally each year is equivalent to around 2% to 5% of the global gross domestic product, or between $800 billion and $2 trillion per year. Meanwhile, data suggests that financial crime generates more than $1.5 trillion in illicit proceeds annually.

Financial organizations have no way to keep up. They’re currently facing increasing regulatory expectations and rising alert volumes, while dealing with a global shortage of experienced investigators. Because of this, they can only prioritize the investigation of a fraction of suspicious financial activities, with significant risk left uninvestigated.

Tangos’ mission is to change that dynamic. It has developed what it says is the world’s most advanced AI-enabled investigative platform, enabling complex financial investigations to be carried out much faster than is possible when relying on humans alone.

The startup, founded by serial entrepreneur Eyal Azoulay last year, brings together a leadership team that possesses a cumulative 75 years of experience in financial crime, sanctions busting and intelligence investigations. Its senior executives include former officials from the U.S. Treasury’s Office of Foreign Assets Control, Israel’s national security community and also AI infrastructure experts with plenty of experience in building autonomous systems.

Whereas traditional compliance tools focus on identifying risk by analyzing financial activity, Tangos automates the investigative work into the suspicious activity it discovers. It merges multiple domain-specific AI models with structured investigative workflows and expert-trained reasoning systems in order to dig up and evaluate evidence, test hypotheses, validate its findings and generate a comprehensive case report that human investigators can then act upon.

Azoulay said Tangos’ system is designed to be ultra-reliable, and for that reason it isn’t made up of a loose swarm of AI agents like other agentic platforms. Instead, the platform is built out of tightly integrated blocks that are specialized to perform different tasks relating to financial investigations, he said.

“Each one runs its own combination of statistical learning and machine learning models,” he explained. “These blocks are orchestrated into playbooks that are tailored to a specific use case, and together they form an autonomous specialist that’s focused on a single investigation type.”

Each playbook was designed with the help of domain experts, Azoulay added, so there’s no improvisation going on in terms of the agent’s reasoning. “They follow the same path that a practitioner in the field would recognize and defend,” he said. “Models operate both inside individual blocks and across the orchestration that connects them, so nothing runs open-ended.”

Tangos’ agents are designed to work top down, so they begin with a business question and then come up with various hypotheses that might answer it, before testing each one against the source data so it can either be confirmed or discarded. This design means that all of its agents’ findings can be traced back to a source, with an audit trail that humans can review and reproduce.

Yaniv Stern of Red Dot Capital said this is a game changer, because though financial organizations have advanced systems in place for detecting suspicious transactions and activities, they don’t have any easy way to investigate what they find. “Investigations remain one of the most resource-intensive functions across compliance and risk organizations,” he said.

Azoulay said AI is ideally suited to these kinds of investigations. These days, financial crime is more of a network problem than anything else, and the complexity of those networks is what exceeds the capacity of human investigators.

“Detection was never the bottleneck, and institutions are drowning in alerts,” he explained. “The bottleneck is what happens after the flag, namely the investigation itself. These investigations are really just network problems, because money moves through seams, across counterparties, jurisdictions and languages faster than any team of human analysts can trace by hand.”

Using Tangos’ platform, customers can immediately launch investigations into any suspicious transactions and compliance alerts they discover, analyze sanctions, anti-money laundering and fraud risks and more, Azoulay said. The platform’s autonomous agents will get started on the grunt work to identify and map out beneficial ownership structures and the hidden relationships between different entities and counterparties.

If they discover evidence of wrongdoing, they will then attempt to validate those findings through multiple sources to strengthen any legal cases that may arise. Once an investigation is completed, it will generate a “regulator-ready” report with a complete audit trail, so authorities can trace everything back to the source and make a decision on what action should follow.

Of course, if a financial organization decides it needs to take legal action after conducting an investigation, it had better make sure that those reports are absolutely watertight in terms of their credibility and correctness. That’s why Tangos’ reports are built to the same evidentiary standards as any law enforcement agency, Azoulay said. “They’re source-traced, reproducible and structured in the way an examiner expects to receive them,” he continued. “The audit trail means any findings can be defended, not just asserted.”

That said, Azoulay was quick to point out that it’s not Tangos that decides how much legal weight a report might have in a specific proceeding. It’s the institutions themselves that must decide.

“Tangos does the investigation and constructs the evidence, and the bank, regulator or legal authority decides if it’s going to be used in any legal action,” he said. “What we guarantee is that when they open the file, nothing in it is unsupported.”

Azoulay would not reveal the names of any of the startup’s customers, but said its platform has already been put to work by a number of major financial institutions and intelligence agencies investigating high-stakes financial crimes, sanctions-busting rings and national security threats.

“We believe autonomous AI will transform how investigations are conducted, and Tangos is uniquely positioned to lead that transition,” Stern added.

Image: Tangos