Anxiety mounts over DOGE’s rapid introduction of AI at regulatory agencies

As experts at a US Securities and Exchange Commission (SEC) conference on artificial intelligence (AI) convened in late March to warn of its risks, the Department of Government Efficiency (DOGE) was already moving through the agency and applying the advanced technology. AI tools have been mooted as a way to identify rules to cut and create efficiencies at slimmed down regulatory agencies.

The SEC was the latest in a series of financial regulators targeted by DOGE for downsizing and deregulating. While public comment has been muted, and some chief executives have lauded the effort to cut regulations, DOGEʼs actions have so far generated anxiety among risk managers at banks and finance firms, who are unlikely to be reassured by the surprise introduction of AI.

Risk and compliance managers may be avoiding public comment but privately they are expressing concern over how to manage the disruption at the SEC, which oversees the $50 trillion capital market, and the banking agencies that monitor and protect trillions more in deposits.

“Large numbers of private actors — attorneys, investment bankers, accountants, and entrepreneurs — can expect their ability to pursue transactions, implement business plans, offer new products, or respond to new developments will be impeded if the SEC’s staff is greatly reduced,” said a group of professors of securities law calling themselves the Shadow SEC, in a statement. 

Similar uncertainty was created at the Federal Deposit Insurance Corp (FDIC), as the insurer of $19 trillion in US bank deposits, after 6,000 jobs were removed from the already understaffed agency. In a report released in late March, FDIC inspector general Jennifer Fain said it might be unable to carry t critical “safety and soundness” examinations.

“With fewer examiners but the same responsibility to conduct statutorily required exams in 2025, it may be difficult for the FDIC to complete these examinations by the end of the year,” she said. With the FDIC still reeling from the largest bank bailouts in history in late 2023, she added, “the full, long-term effect of the restructuring and reshaping of the FDIC is unknown, as these activities are ongoing”.

Legal experts, meanwhile, are advising firms to maintain robust compliance since rules and regulations remain largely unchanged despite the whirlwind hitting their regulators.

Uncertainty

The changes have come so fast that risk experts are struggling to keep up with developments since the current uncertainty and lack of transparency makes forecasts virtually impossible. The AI-first transformation is taking place in every corner of government, at pace.

“What took centuries in the colonial era is now unfolding at lightning speed in mere days through digital means,” Allison Stranger, a technology and government fellow at Harvard’s Ash Center for Democratic Governance and Innovation, said in a blog. Market volatility and political uncertainty have led to instability. 

Amid such rapid change, “audit, risk and compliance professionals in the financial services industry will need to remain particularly attuned to shifts in these risks”, Richard Chambers and Associates said in its Audit Beacon advisory.

The FDIC’s Fain said her agency will have fewer resources to provide guidance to banks, which could raise the importance of in-house and third-party risk and compliance experts. Furthermore, according to the Shadow SEC, the SEC may be unable to provide timely responses to queries or to correct deficiencies in regulatory filings. That means “greater risk and liability at a later point when litigation may be brought by either the SEC or private plaintiffs”.

Pushback against DOGE is increasing, and courts have ordered agencies to reinstate large numbers of axed workers, including the entire 1,500 staff of the Consumer Financial Protection Bureau. However, if or when Elon Musk departs his DOGE role, many of his changes will be irreversible such as job cuts .

Uncertainty will continue for months due to scores of legal challenges over alleged invasions of privacy, cyber security risk, lack of constitutional authority, and violations of labour union contract, civil service rules and the Administrative Procedure Act that governs agency processes.  Enforcements and examinations are continuing at agencies, although in less predictable ways.  

State and local authorities remain on the job, as well, along with industry self-regulators such as the Financial Industry Regulatory Authority (FINRA). While the Trump administration has pulled back from enforcement of bribery and money laundering, sanctions remain in place and the Office of Foreign Assets Control (OFAC) “strongly encourages parties to exercise due diligence when their business activities may touch on an OFAC-administered programme”.

Quick wins

DOGE was launched with a dual mission of reducing both regulation and government expenditure, however it quickly focused on cost cutting as its primary goal. It has opted for “the quick wins” that AI can deliver, according to Joseph Haslag, professor of economics at the University of Missouri and a former economist at the Dallas Federal Reserve Bank.

“In terms of expediency, the dollar metric almost had to win out,” Haslag told Compliance Corylated. “In short-term versus long-term thinking, it calls for that; regulations take a good amount of time [to change] .”

Proper use of AI first necessitates fine-tuning the prompts that guide the algorithms, Haslag added. “Without those, I donʼt think AI going to help and it could hurt if you just start applying this without a clear objective.”

AI experts at the SEC conference echoed the need to establish careful controls when applying AI, to avoid unintended consequences. Deep Ratna Srivastav, head of AI and digital transformation at Franklin Templeton Investments, said: “Itʼs a slow, deliberate and very long-term process, and that is why you see those questions come up, because it cannot happen overnight.”

Because AI works across multiple data sets and systems, oversight is critical, Srivastav said. “Small control teams are doing it with a lot of oversight. You are taking lots of risks. The risks of a large language model are already out there and known in the market.” These vulnerabilities need to be “mitigated, with controls in place” from the start.

Haslag called AI “unequivocally a great tool” that could be highly effective over time in eliminating redundant regulations and making government more effective. But misapplied “it may not have anything to do with the underlying frictions that are causing inefficiencies to exist”. Meanwhile, the risk of privacy invasions, cyber attacks, mistakes and other unintended effects, has led many to wonder if the reward makes it worthwhile, he added.