Non-stop agentic AI action has teams assembling to face the threat

Agents: you can’t live with ‘em, you can’t live without ‘em. So it seems in today’s tech world, which is chewing away at the problem of how to make maximal use of agentic capabilities without letting bots run wild in the spreadsheets. Autonomous risk is real – but so are the heroes who can stop it, with authentication tools including biometrics, continuous monitoring, device intelligence, and more.

Human-agent team star in new Sphinx compliance platform, Frontline

Sphinx Frontline is the new product from Sphinx, an AI compliance startup that raised $7.1 million in seed funding led by Cherry Ventures earlier this year, with participation from Y Combinator, Rebel Fund, Deel Ventures and Singularity Capital.

A release calls Sphinx Frontline “the world’s first AI-native BPO for financial crime compliance.”

Frontline hires compliance analysts and embeds them into customer workflows, where they work cases alongside Sphinx’s browser-native AI compliance agents – and continuously retrain those agents based on new data. If an analyst sees the agent disagree with how a case was historically resolved, they can “investigate, uncover an undocumented policy rule, update the SOP, retrain the agent, and validate it against thousands of historical cases before deploying.”

The result, says Sphinx, is “a compounding feedback loop that most institutions take months to build, delivered on day one.”

“Banks don’t have to ‘implement AI.’ They hire a team.”

The product is already live with regulated public financial institutions, which report an 80 percent reduction in average case review times. Alexandre Berkovic, CEO of Sphinx, believes compliance has never been a problem with software exclusively. “Software is starting to eat services, but human expertise is becoming more important, not less,” he says. “Frontline is the new bar – not pure software, not pure outsourcing, but AI-native compliance teams where humans and agents continuously learn from each other.”

GBG launches GBG for Agents

GBG PIc has launched its GBG for Agents suite to service the market for agentic AI decisioning and governance. A post on LinkedIn says the first product to roll out is GBG Reach, a decisioning layer that makes Loqate’s verification capabilities available to AI agents and agentic workflows.

“The way location and identity verification gets consumed is changing,” says Matt Furneaux, GBG’s global director of location product and data strategy. “It’s not just humans calling APIs anymore. Agents are selecting providers, orchestrating checks, making routing decisions. We want to make sure that when an agent needs trusted address, identity, or contact verification – it finds us.”

“This isn’t a rebrand or a wrapper. It’s new infrastructure, purpose-built for this moment, and we shipped it at pace.”

Agentic AI deployment ‘perilous, vertiginous, catastrophic’

What, exactly, is this moment? In an opinion piece on the CSO blog, Sunil Gentyala suggests it is the very edge of the apocalypse.

“We find ourselves teetering upon a precipice of our own unwitting construction, and the vertiginous depth of our collective negligence ought to give every security practitioner profound pause,” Gentyala writes. “In our headlong rush to deploy AI agents across enterprise environments, we have erected an infrastructure so thoroughly unfortified that it beggars belief.”

Anthropic’s Model Contest Protocol (MCP), unveiled in November 2024 as the connective tissue binding large language models to external tools, has “proliferated with breathtaking celerity.”

“What has conspicuously failed to keep pace is any semblance of security discipline. The chasm between adoption velocity and security maturation grows more perilous with each passing deployment.”

“Your AI infrastructure represents either an invaluable asset or a catastrophic liability. The adversary has rendered their assessment with cold-eyed clarity. The window for meaningful action contracts with each passing week, and the cost of inaction compounds exponentially. The framework exists. The architecture is implementable. What remains is organizational will.

“Have you made yours?”

Article Topics

AI agents  |  cybersecurity  |  financial services  |  GBG  |  identity security  |  Sphinx