Page 26 - gs260102
P. 26
Cover Story continued
ignoring age restrictions or enabling purchases prohibited
in certain jurisdictions. Step-up authorizations for pricing
thresholds, purchase frequency or delivery locations,
along with robust monitoring and "circuit breakers" that
can execute transactions continuously, mak- intervene when agents make errors or are manipulated,
ing errors or exploitation more damaging can help mitigate these risks, he added.
than traditional human-driven fraud. Risk
managers cite deep fakes, synthetic identities, Rather than becoming complacent or affirming controls
fake merchant onboarding and rapid bust- are already in place, Leach stated, businesses should
out schemes as growing concerns. Mitigation take AI agents seriously because they have changed the
strategies include transaction velocity limits, consumer variable.
behavioral monitoring, real-time anomaly
detection, and circuit breakers that halt agent Service providers need to re-evaluate every aspect of
activity when thresholds are exceeded. security and fraud prevention as we evolve beyond static,
human-operated transactions, he added.
4. Policy and jurisdictional violations
Non-human authentication
Agents that transact across borders or catego- Leach recommended leveraging existing security frame-
ries may inadvertently violate legal or policy works and advanced tokenization, such as Visa's Trusted
constraints, such as age restrictions, pro- Agent Protocol and Mastercard's Agent Pay, to verify and
hibited goods or regional purchasing rules. authenticate AI agents. "We need to think of agents as an-
Experts emphasize embedding compliance other type of non-human identity (NHI)," he said.
logic directly into agent permissions. Step-
up authentication for sensitive transactions, "Bolting additional trust layers onto agentic AI communi-
geographic and category-based restrictions, cation that are not inherent, such as encrypted authentica-
and continuous audit trails are increasingly tion, will help distinguish 'good agents' from 'bad bots.'"
viewed as baseline requirements rather than
optional safeguards.
5. Infrastructure and protocol maturity 2026 Events
New communication protocols enabling Nacha’s Smarter Faster Payments 2026
agent-to-agent and agent-to-system interac-
tion remain in early stages. While tools such April 26 - 29, 2026, San Diego, Ca.
as Model Context Protocol enable powerful
integrations, experts warn that early stan- IPA’s Innovative Payments Conference
dards may resemble legacy internet protocols April 29 - May 1, Washington, DC
that required decades of layered security to
compensate for original design limitations. SEAA 2026
Institutions are encouraged to treat early
agentic infrastructure as provisional, moni- June 8 - 11, 2026, Miami, Fl.
toring usage closely, isolating failures and
avoiding assumptions of inherent trust. MWAA 2026
July 22 - 23, 2026, Chicago, Il.
6. Governance and human oversight
MAG Payments Conference 2026
Despite automation gains, experts stress that
human judgment remains essential. Agen- Sept. 27 - 30, 2026, Nashville, TN
tic systems should augment, not replace risk
teams, underwriters and compliance officers. Visit www.greensheet.com/datebook.php for
Clear accountability frameworks, escalation more events and a year-at-a-glance event chart
paths and human-in-the-loop controls are or email events@greensheet.com if you have a
critical during early adoption. Organizations event/show you wish to be listed.
that assume existing controls are sufficient
risk repeating mistakes seen during earlier
payment technology rollouts.
26
