The memo dropped late on Monday evening after a weeklong flurry of executive orders from President Donald Trump, and was rescinded less than 48 hours later. But the mandate to pause all federal loans and grants has reverberated across the country, from tiny nonprofits to sprawling health agencies. 

Providers of community services struggled to access government funds and now question what they can rely on going forward. Medical researchers are unclear how grants will be affected. States are grappling with how to plan their budgets.

The directive from the Office of Management and Budget, calling for a freeze on all federal grants, was temporarily blocked by a federal judge and raised questions about the limits of the president’s power. Yet the chaos it unleashed previews the types of fights to come in an administration that has made clear that it plans to reshape the U.S. government and eliminate what it considers to be wasteful spending and policies at odds with its conservative agenda.

“It’s hard for people to plan for the future and maintain current program services if they don’t think there’s going to be funding for those services going forward,” said Susanne Byrne, executive director of the York Street Project in Jersey City, New Jersey. 

Her organization, which offers rental assistance and operates a homeless shelter and a childhood development center, was set to make rent payments coming due on apartments for 45 different families using federal funds from the Department of Housing and Urban Development on Tuesday. The staff found themselves locked out of HUD’s payment portal.

Byrne spent Tuesday on numerous calls with other housing groups and with local representatives of HUD. Not only were the families she works with at risk of eviction if landlords didn’t receive payments by early February, but she worried how her own staffers might fare if the organization had to pare back its activity or lay off workers. 

“If people lose their jobs, how do they support their families themselves? They’re going to end up being people who need the services they used to provide to others,” she said. The group eventually regained access to the portal.

The White House abruptly rescinded the funding freeze memo on Wednesday, a move Press Secretary Karoline Leavitt immediately muddied with a social media post that said the rescission of the memo doesn’t end the pause on government money flows. Judges overseeing two pending legal challenges could rule in the coming days providing more clarity on what — if any — funds could continue to be held up. 

The confusion added to turmoil among federal workers who were already grappling with an executive order last week that banned diversity, equity and inclusion policies from the federal government. The order said the administration would cut off funding for programs that support those efforts, though the exact impact is still unclear.

The Department of Government Efficiency, led by Elon Musk, is trying to find ways to slash more expenses. Trump is offering buyouts to federal employees who were warned that he’s seeking a “more streamlined and flexible workforce.”

Agency anxiety

Already, the planned cutbacks to DEI have raised concerns about funding for medical research programs. The Trump administration sent a memo this week to organizations that have received funding from the Centers for Disease Control and Prevention and ordered them to end “all programs, personnel, activities, or contracts promoting” DEI, according to the document seen by Bloomberg News. 

Georges Benjamin, executive director of the American Public Health Association, said multiple health departments received the memo, and it wasn’t immediately clear how far-reaching the impact could be. He said it could affect the future of a wide range of public health projects, including research into the racial disparities of heart disease and cancer or programs for the disabled or elderly.

“The challenge is we’re allowing people to redefine DEI as not being merit-based when that’s not what it’s about,” Benjamin said. “It targets services to people who are more in need than others.”

The prospect of cuts has led to confusion and anxiety at the National Institute of Health, which invests most of its nearly $48 billion budget in medical research. Almost 83% of that goes toward grants awarded to outside scientists, institutions and medical schools across the US, according to the agency.

Employees have been advised to refrain from promoting certain funding opportunities related to disadvantaged populations so as not to draw attention to them, according to a person familiar with the matter, who asked not to be named speaking about internal discussions. NIH staffers have been warned against making social media posts that could put a target on their offices, the person said.

The NIH said in a statement that the Health and Human Services Department has issued a pause on mass communications and public appearances that aren’t related to emergencies to allow the new team “to set up a process for review and prioritization.” Clinical trials are continuing, though no new studies are being launched.

Within the Commerce Department, officials in charge of a $700 million effort to remake economically depleted cities into hubs of technological innovation scrambled in the wake of this week’s funding freeze announcement. The so-called Tech Hubs project was on a spreadsheet circulated by OMB that listed grants under scrutiny, instigating confusion among recipients of the grant funding, as well as former and current officials who helped to put the plan together.

The Tech Hubs initiative, funded by the Chips Act and a subsequent defense spending bill, poured hundreds of millions of dollars into technological projects in more than a dozen states, including red states like Indiana, Montana, South Carolina, Georgia and Ohio. Many of the projects focus on industries that the Trump administration has identified as key national security priorities, such as semiconductor manufacturing, quantum computing and critical mineral processing. 

But some of the hubs have programs aimed at including more women and minorities in the workforce, while others include “equity” or “climate resilience” in their missions.

So far, the tech hubs haven’t received any guidance from the Commerce Department, said spokespeople for six of the hubs. People within the hubs are setting up meetings with their lawmakers and trading texts with staffers on Capitol Hill, seeking a full-throated commitment that they won’t lose funding. The Commerce Department didn’t immediately return a request for comment.

‘Unnecessary disruption’

At the local government level, the questions over federal aid introduce new uncertainty at a time many states are preparing their budgets for fiscal 2026, said Lucy Dadayan, principal research associate with the Urban-Brookings Tax Policy Center at the Urban Institute.

This has “caused unnecessary disruption and confusion, as well as considerable anxiety regarding the amount of federal funds that will be allocated to the states,” she said.

Local organizations are preparing for more shocks. The childcare industry, for one, is bracing for potential changes that could upend businesses already on the brink of collapse. Federal dollars are states’ main funding source for subsidies that help low-income families afford child care. Some 1.8 million children receive them each year, according to the U.S. Government Accountability Office.

Easterseals, a 106-year-old nonprofit that provides services to people with disabilities, senior citizens, veterans and their families, was concerned about funding troubles when its payment system portal shut off hours before the OMB memo was released, said Kendra Davenport, CEO of its national office. The group’s affiliates use this portal to access federal grants to cover operational costs and payroll. 

“It really was foreboding not knowing how quickly we were going to have to shut down programs that are 100% funded with federal funds and how quickly we would have to determine how many people we were going to lay off,” Davenport said.

After hearing about the freeze, Easterseals assessed whether it could maintain programs without federal funding but knew that most nonprofits, including its own, operate on tight margins. Most affiliates are so focused on programmatic work that they don’t have time to raise private funding that would fill gaps, Davenport said.

As the national office looks at future revenue, staff are scrutinizing nonessential expenses that can be taken out of the budget and put into reserves. 

“We now know we are operating in a very different federal funding climate,” Davenport said. “So we have been very careful.”

Accounting artificial intelligence leaders have deemed Chinese large language model DeepSeek as roughly on par with other general models when it comes to knowledge, questions and tasks relevant to the profession. 

DeepSeek’s latest R1 model was released to the world last week to much fanfare by producing performance comparable to massive models like Claude or ChatGPT but at a fraction of the cost — something that blindsided other players in the AI field who hadn’t believed it was possible. In the short time since it exploded on the world stage, industry observers have had to rethink their priors, as the event has shown that one does not have to be a gigantic corporation like Microsoft (whose $3 trillion market cap is bigger than the entire GDP of Italy) to produce quality AI models. 

Generative AI models are not usually known for their math skills, let alone accounting, as the probabilistic nature of their outputs makes precision difficult. While theoretically someone could apply public models like ChatGPT to an accounting problem, the possibility the AI could make up information from whole cloth makes it a risky proposition. This does not, however, mean they’re completely incapable of performing accounting tasks, just not as good as a more specialized solution. 

Alexandr Vasilyev – stock.adobe.

Jeff Seibert, CEO of accounting automation solutions provider Digits, tested DeepSeek against OpenAI’s ChatGPT, Anthropic’s Claude and Meta’s Llama and found its ability to perform accounting tasks and answer relevant questions to be roughly on par with the others. He asked the AIs to classify 1000 transactions, given the description, dollar amount and a chart of accounts. ChatGPT was correct 61-65% of the time depending on the specific version; DeepSeek was correct 59.90% of the time; Llama was correct 48% of the time; and Claude was correct 43.4% of the time. The correct category was in the top 3 suggestions 75-78% of the time for ChatGPT (depending on version), 69.67% for DeepSeek, 61% of the time for Llama and 56% of the time for Claude. 

ChatGPT created a fake category 0.10-4% of the time (depending on version), Llama did so 0.2% of the time, Claude 2.8% of the time, and DeepSeek 0.39% of the time. As far as speed, it answered queries faster than GPT-o1 mini, GPT-o1, and Llama. It was slower than GPT-4o, GPT 4 Turbo and Claude. 

Seibert, in a LinkedIn post outlining his experiment, noted that even if it doesn’t outdo the other models in all areas, the fact that it was made at a fraction of their cost is quite impressive. 

“If their claims around training cost are accurate, this represents a massive breakthrough in model efficiency and sets the new bar for open source AI performance,” he wrote. 

Daniel Shorstein, president of technology solutions advisory firm James Moore Digital, also put DeepSeek through its paces by asking multiple choice questions covering a range of accounting topics, adding he tried to keep them just difficult enough that it would trip up a less than highly intelligent LLM. He used the same test as the one he made to evaluate Llama, Claude, ChatGPT and other models against each other. He found its results to be inconsistent. 

He illustrated with how it reacted to a question on segregation of duties: 

Question: “There are three employees in the accounting department: payroll clerk, accounts payable clerk, and accounts receivable clerk. Which one of these employees should not make the daily deposit? A. payroll clerk B. account payable clerk C. accounts receivable clerk D. none (any can make the deposit)”

DeepSeek, like other models lately, is equipped to not only provide an answer but reveal some of its internal reasoning in how it got to the answer. Shorstein noted that, internally, it actually got the correct answer after a long chain of reasoning where it first recalled general principles of segregation of duties, thought of an ideal setup, thought of possible exceptions and special circumstances, went back to the main point of segregation of duties and ultimately determined the answer was C, the accounts receivable clerk, which Shorstein said was the correct answer. 

“But its final answer: ‘The correct answer is A. payroll clerk,'” he said in a message.

Meanwhile, Hitendra Patil, CEO of accounting tech consultancy Accountaneur, said DeepSeek gives more detailed answers compared to ChatGPT, and also appears to have been built to not only answer the direct question asked, but to actually pre-empt the likely follow-up question that the user may ask after the first question and provides answers to such pre-empted follow-up questions. He added that it also goes over its reasoning when discussing math questions versus other models which simply give an answer.

For example, he asked the model what is 343 multiplied by 741. It broke down the problem using the distributive property to simplify the multiplication, then added the results together to get 254,163. 

At the same time, he noted that DeepSeek does not browse the Internet unless specifically asked, and so some of the answers it gave him about tax law were somewhat dated, versus ChatGPT which more or less gave the latest information. Overall, he said that it seems slightly behind ChatGPT for now, despite rumors that it had been trained similarly.  

“There is no verifiable proof … but DeepSeek has been/is being trained on similar underlying data that ChatGPT was/is being trained on, albeit it seems to be lagging behind ChatGPT,” he said in an email.

Table disservice; down on the farm; a barebones job; and other highlights of recent tax cases.

Miami: Tax preparer Juan Mendieta has been sentenced to 57 months in prison after pleading guilty in October to one count of criminal conspiracy and four counts of aiding and assisting the preparation of false returns.

Beginning in 2019 and continuing through at least 2023, Mendieta conspired to prepare fraudulent returns for clients by using false business losses and expenses. These false items resulted in inflated federal refunds.

For multiple clients, he prepared two sets of returns. One set directed certain federal refunds to Mendieta’s clients; Mendieta provided this set to clients and told them he would file these returns with the IRS. Instead, he filed a second set of returns that directed even greater refunds to bank accounts that he and a conspirator controlled.

The IRS has identified at least 29 tax filings that fraudulently inflated refunds, which Mendieta filed on behalf of at least 13 clients. The IRS is entitled to more than $11 million in restitution.

Boston: Restaurateur John Drivas, of Hampton, New Hampshire, has been sentenced to a year and a day in prison, to be followed by a year of supervised release, for defrauding the IRS of federal employment taxes and the Massachusetts Department of Revenue of state meals taxes.

Drivas, who pleaded guilty in September, owned and operated three restaurants in Salem and Peabody, Massachusetts, and in Seabrook, New Hampshire. He was the sole shareholder of the Salem restaurant until he sold it to an employee in 2022, the 100% owner of the Peabody restaurant with his wife and the 52% owner of the Seabrook restaurant with his children.

From at least January 2017 to June 2022, Drivas paid under-the-table wages of $1,496,417 to multiple restaurant employees and did not report those wages to the IRS or pay employment taxes on them, causing more than $439,000 in employment tax losses.

He also collected the state and local meals taxes paid by restaurant customers, which he failed to pay over to the state: In Massachusetts, owners and operators of restaurants and bars are required to collect 6.25% sales taxes on meals. Salem and Peabody also require restaurants and bars to collect an additional 0.75% local option meals excise tax. Although Drivas collected the taxes from restaurant customers, he intentionally withheld $1,596,775 of those taxes from monthly reports and payments owed to Massachusetts.

Drivas was also ordered to pay restitution of $1,596,775 to the state and $439,341 to the IRS, in addition to a $20,000 fine. 

Los Angeles: Area resident Kevin J. Gregory has pleaded guilty to seeking more than $65 million by falsely claiming that his non-existent farming business was entitled to pandemic-related tax credits.

From November 2020 to April 2022, he made false claims to the IRS for the payment of nearly $65.4 million in tax refunds for a purported Beverly Hills-based farming-and-transportation company named Elijah USA Farm Holdings. The IRS issued a portion of the refunds Gregory claimed, and he used that portion — more than $2.7 million — for personal expenses.

Specifically, in January 2022 Gregory made a false claim to the IRS for the payment of a tax refund of $23,877,620, which he submitted as part of Elijah Farm’s quarterly federal return. He claimed that Elijah Farm employed 33 people, paid nearly $1.6 million in quarterly wages, had deposited nearly $18 million in federal taxes and was entitled to nearly $6.5 million in COVID-relief tax credits. In fact, Elijah Farm had no employees and paid wages to no one and had not made federal tax deposits in the amounts stated.

Sentencing is May 16. Gregory, who’s been in federal custody since May 2023, faces up to five years in prison.

Jacksonville, Florida: Jose Molina-Herrera, of Honduras, has been sentenced to 27 months in prison for conspiracy to commit wire fraud and conspiracy to defraud the U.S. for the purpose of impeding the IRS.

Between 2019 and 2020, Molina-Herrera conspired to facilitate payment of construction workers off the books to avoid paying payroll taxes and premiums for workers’ compensation insurance. Construction contractors and subcontractors entered arrangements with the conspirators through which All National Remodeling, a shell company formed by Molina-Herrera, facilitated both the distribution of proof of insurance and the payment of workers with cash.

In exchange for 6% to 8% of the contractors’ and subcontractors’ payroll, Molina-Herrera and others caused the distribution of certificates of liability insurance in the name of All National, which contractors and subcontractors then used as nominal proof that workers were supposedly insured. In reality, All National Remodeling’s insurance policy was issued based on a fraudulent application that never disclosed that contractors and subcontractors would be employing workers who were ostensibly insured under the shell company’s barebones insurance policy. The insurance company was defrauded of more than $2.2 million.

Molina-Herrera and others also facilitated deposit of checks into the shell company’s bank accounts as well as the withdrawal of cash to be paid to workers, all without withholding, or paying over, payroll taxes to the IRS. Through these arrangements with the conspirators, the construction contractors and subcontractors could disclaim responsibility for withholding and paying payroll taxes to the IRS or ensuring that the workers were legally authorized to work in the United States. By facilitating payments to workers of more than $14 million without payroll taxes being withheld, Molina-Herrera and his co-conspirators caused the U.S. Treasury to lose more than $3.5 million in tax receipts.

Molina-Herrera, who pleaded guilty in November, was also ordered to forfeit $867,005, the proceeds of the wire fraud, and was ordered to pay $3,558,579.42 in restitution to the IRS. One co-conspirator, Oscar Molina-Avila, was previously sentenced to 52 months in prison for his role in the scheme.

Agate, Colorado: Businesswoman Shandel Arkadie has pleaded guilty to not paying employment taxes.

Arkadie operated Alternative Choice Home Care Nursing and was responsible for withholding Social Security, Medicare and income taxes from employees’ wages and paying those funds over to the IRS each quarter. She was also responsible for paying over Alternative’s portion of Social Security and Medicare taxes.

Between January 2015 and December 2020, the company withheld more than $1 million from employees’ wages but did not pay the funds over to the IRS or file the quarterly returns. The company also owed some $500,000 in Social Security and Medicare taxes that Arkadie did not pay.

In total, she caused a tax loss to the IRS of some $1.5 million.

Sentencing is May 15. She faces a maximum of five years in prison, a period of supervised release, restitution and monetary penalties. 

Cogan Station, Pennsylvania: Businessman James Michael Barr has been sentenced to time served plus two years of probation, including 10 months of home confinement, for failing to pay employment taxes owed by his construction company.

Barr pleaded guilty in July to failing to account for and pay over employment taxes owed by Barr Construction from 2017 through 2020. In addition to a normal paycheck from which taxes were withheld, Barr also paid his employees in cash and did not withhold federal taxes from the cash payroll or remit taxes to the IRS.

The sentence also imposed a $5,000 fine and required Barr to make $337,000 in restitution to the IRS, plus penalties.