Senate Democrats are asking for more information on the Social Security Administration’s use of artificial intelligence on its phone lines following reports that an anti-fraud tool degraded service over the spring and that a new AI bot on the phone lines is causing service issues for callers now.
“These AI programs, which the agency deployed with little consultation with Congress, advocates, or other key stakeholders, appear to have been developed in haste and represent a troubling pattern that if continued, would significantly impede Americans’ ability to access their Social Security and Supplemental Security Income (SSI) benefits,” wrote Sens. Ron Wyden, D-Ore., Elizabeth Warren, D-Mass., Kirsten Gillibrand, D-N.Y., and Bernie Sanders, I-Vt.
SSA changed its policy of holding claims to run the anti-fraud checks in mid-May after Nextgov/FCW reported that the tool slowed retirement claim processing while only finding two claims out of over 110,000 with a high probability of being fraudulent. That fraud check was installed after DOGE and White House leaders repeated false claims about fraud on SSA phone lines.
“Only after significant public outcry did SSA decide to pare back its policy,” the letter — dated June 24 but publicly released July 1 — reads. “To date, SSA has yet to provide us any evidence or analysis demonstrating the need to change its policy from its pre-March anti-fraud measures.”
The letter also points to a new, AI-based chatbot on SSA’s national 1-800 number, which the agency plans to extend to field office numbers as well. News reports have detailed issues with the bot, like it not putting callers through to agents, as well as long wait times and disconnections before callers reach a real person.
“We have also received reports of service issues from constituents, advocates and key stakeholders, such as the chatbot responding to a different question than what the caller had asked or incorrectly disconnecting the call before the caller’s questions had been answered,” the letter reads.
The senators are asking for more information on the phone chatbot, including how SSA is evaluating it, as well as what AI the agency is planning to deploy this year.
But an agency spokesperson told Nextgov/FCW that SSA has been fully transparent — including via congressional briefings — around the use of AI on the 800 number, which is listed in the agency’s publicly available AI inventory.
“The AI behind the question and answer bots simply matches the words the customer speaks with frequently asked questions. It does not generate the final response to callers questions,” said SSA spokesperson Stephen McGraw, noting that the agency performs daily analytics and effectiveness can be impacted by factors like background noise or the quality of the connection.
Last week, the agency’s commissioner, Frank Bisignano, told lawmakers that he has a “technology agenda” for the agency as it sheds thousands of employees. The group of four senators appear skeptical.
“We highlight these recent cases to illustrate our concerns that the agency, under this administration, appears to be ill-equipped to develop and deploy AI,” the letter reads. “If SSA cannot learn from its mistakes in incorporating AI into basic telephone services for beneficiaries, we are concerned that SSA will make even bigger mistakes in incorporating AI into higher-risk tasks, particularly in roles that could jeopardize Americans’ financial security.”
