Sigma360

Deciding whether a named person is an article's subject, a witness or a namesake is the work the models do, across an adverse media corpus stated at more than 225 million articles from 730,000 publishers. List screening would survive without inference, but it supports the adverse media engine and the AI360 investigator agent rather than carrying the product.

The AI Investigator Agent can close easily cleared alerts on its own or give explainable recommendations an analyst accepts, rejects or modifies, so the institution chooses whether a person sits in the execution path. Its confidence threshold for automatic closure, and any category the agent may never close, are not published.

An override dashboard tracks how often analysts reject the agent over time, a live monitor a model risk function can use, and Yanez Compliance is named as independent validator without a published method or result. Automation of 93 percent of manual alert clearing and 90 percent less match review come with no miss rate, though fewer false positives can mean detecting less.

Named quotes from Rippling, Paypercut and Solatrix carry no outcome figure, while the figures sit on anonymized cases, such as a global payments leader saving $1 million a year. Company figures add fivefold growth in two years and net revenue retention above 140 percent, and Chartis named it a category leader in adverse media monitoring in 2025 and 2026.

Analyst overrides are a valuable training signal, and whether screening lists, adjudications or override histories improve models serving other institutions is not stated. Retention of prompts and summaries, and whether a model provider receives screened subject data, are unstated too.

Cookies, newsletters and registered users fill the privacy policy, which moves website data from the EU to the US under standard contractual clauses. The people whose names and adverse coverage populate the risk database, how long their profiles last and how a wrongly attributed article is corrected are never addressed, and a line about voter information reads as template text.

ISO 27001:2022 certification and a SOC 2 Type II report appear in the footer, a Vanta trust center runs at the firm's own subdomain, and a disclosure policy gives researchers a reporting route. That footer calls SOC 2 a certification, though it is an attestation over a period, and certifier, scope and report period are not on open pages.

Registered status was never part of the original non credit rating business, and the current platform states no authorization or position on examiner review of automatic alert closure. A founder with US Treasury and National Counterterrorism Center experience brings expertise rather than standing.

Matching names across 124 languages does not fail evenly, so the odds of a wrong match to an adversely reported namesake, or a wrong clearance, depend on where a name comes from. Country risk ratings are sold beside entity screening, and whether a jurisdiction rating moves an individual's result is left open, with no per language accuracy or fairness testing published.

When the agent closes alerts on its own, the decision not to investigate is the model's under the institution's configuration, and nothing allocates responsibility for a wrongly closed alert. Terms of service, warranties and service levels are unpublished, and the privacy policy puts the risk of sending information on the user.

Large language models write the adverse media summaries and entity reports, while entity resolution and match scoring are presented as in house, yet no model provider, hosting arrangement or sub processor list is named. Whether summarization sends screened names to an external service is not said.

Since June 2026 Sigma360 has powered screening and monitoring inside Rippling's global payments infrastructure, nine months after selection, and Smile ID embeds it in a white labeled KYC product. Spheros and Fideo Intelligence partnerships add business and identity data, yet no core banking, onboarding or case management connector is named.

Hosting is described as a secure cloud environment with a public status page, and no provider, region or residency commitment is named. Institutions outside the United States with confidential customer lists are among the buyers.

A free trial of the full enterprise platform follows a short onboarding session, but no rate card, tier or billing unit appears on any solution page or in the comparison guide. Screened names, monitored entities and seats would produce very different bills at scale.

Global banks to early stage fintechs are served, with named customers in payroll and payments at Paypercut and Oyster, in banking at Barclays and in asset management at Magnetar Capital. Corporates outside finance such as Trafigura and Enverus also buy, so finance is a major vertical rather than the entire market.