Seems a bit DOGE-y: Protecting Federal Worker Rights in an Era of AI-Driven Employment Decisions

In the last month and a half, the Trump Administration has systematically gutted the federal workforce. However, the Administration’s latest assault brings federal workers a new concern: that they will be fired by an algorithm. Last Wednesday, WIRED reported that DOGE engineers are modifying a decades-old program called AutoRIF. AutoRIF was designed to assist human decision-makers in determining which agency employees will be let go in periods of mass layoffs, called Reductions in Force (RIF). The next day, the Administration put out a memorandum ordering every agency to submit a RIF Plan.  

It is not clear how AutoRIF has been modified or whether AI is involved in the RIF mandate (through AutoRIF or independently). However, fears of AI-driven mass-firings of federal workers are not unfounded. Elon Musk and the Trump Administration have made no secret of their affection for the dodgy technology and their intentions to use it to make budget cuts. And, in fact, they have already tried adding AI to workforce decisions. On Saturday, January 22, a DOGE email ordered federal workers to either account for their time or face termination. (Despite attempts to backpedal, DOGE sent essentially the same email again just last weekend). It was later reported that DOGE intended to feed responses into a large language model (LLM) that would be tasked with deciding whether the worker is necessary. 

All of these actions have raised questions about federal employee rights and how AI can and cannot be used in termination decisions. This blog post provides an overview of a federal worker’s rights when faced with termination, the problems with using automated decision-making tools to conduct terminations, and how those tools may violate worker’s rights.  

What rights do federal workers have?

The federal workforce is critical to maintaining government services, benefits, and functions, across administrations. Over 3 million Americans from across the U.S. are employed by the federal government. When faced with individual or group termination, federal workers have constitutional, statutory, regulatory, and contractual rights. If the government wants to remove federal workers using automated tools, it must still respect these rights and follow the proper procedures. 

Protection 1: For-Cause Employment Status 

The scope of a federal worker’s protections and rights depends on their position, worker classification, and tenure. Unlike those who are privately employed, most civil service workers are for-cause employees. This means that the worker cannot be removed from their position at the will of their employer when their removal is not part of a larger Reduction in Force (take a look at Protection 2 below). Instead, they can only be removed for a recognized legitimate reason or “cause”. A “cause” may include poor performance, misconduct by the worker, or economic pressures. A for-cause federal worker who believes they were not terminated for legitimate cause may go to the Office of Special Counsel to file a complaint.  

Protection 2: Reduction In Force Regulations and Procedures 

Reductions in Force action, known as a RIF action, is the term used to refer to federal government layoffs. RIF actions happen for organizational reasons like a lack of work, restructuring the agency, or a shortage of funds. Presidents and agencies have the authority and discretion to conduct RIF actions under 5 U.S.C. §§ 3501—3504. When an agency conducts a RIF action, it must follow the RIF regulations issued by the Office of Personnel Management. These regulations determine whether a worker will keep their position and whether the worker has a right to a different position if their position is terminated as part of a RIF action.  

Most federal civilian workers have rights under the laws and regulations that govern RIF actions. For example, a RIF action cannot be used to remove or demote a worker for individual reasons like conduct or performance. Instead, RIF actions must happen for organizational reasons. A worker terminated as part of an RIF action must be given advance notice and termination benefits.  

RIF actions have already begun at agencies like the Social Security Administration. Unfortunately, some career civil servants terminated in recent RIF actions allege that proper procedures were not followed. A terminated worker who believes that the agency failed to follow RIF procedures may appeal their termination to the U.S. Merit Systems Protection Board.  

Protection 3: Prohibited Personnel Practices 

In addition to violations of RIF procedures, as noted above, the Merit Systems Protection Board (MSPB) also hears appeals from workers who have suffered a “prohibited personnel practice,” or PPP. Generally, federal workers are protected from PPPs including discrimination, retaliation, or interference with competition for a position. These protections are statutory under 5 U.S.C. § 2302(b). Another part of that law protects workers from actions that are arbitrary, coercive for political purposes, or otherwise violate the merit system principles that government employment supports. Workers who believe they have suffered a PPP or that the merit system principles have been violated should contact their Office of Special Counsel.  

Protection 4: Civil Rights Act Title VII 

Federal workers, nearly 20% of whom are Black, have the right to equal employment opportunities and the right to be free from discrimination based on race, color, religion, sex (including gender, gender identity, pregnancy, and sexual orientation), and national origin. Title VII of the Civil Rights Act of 1964 protects all workers from discrimination in employment decisions like hiring and firing.  

Protection 5: Due Process under the Fifth Amendment 

Federal non-probationary workers have due process rights through the Fifth Amendment to the Constitution as well as civil service law. If a worker is being fired in their individual capacity, as opposed to an RIF action, they are entitled to written advance notice, the ability to respond and be represented, a written decision, and the ability to appeal that written decision. Workers affected by an RIF action are entitled to advance notice, appeals, and termination benefits.  

Protection 6: Collective Bargaining Agreements 

Nearly 36% of the public sector workforce is unionized. Since the Inauguration, federal labor unions have seen a surge in membership. Labor unions engage in collective bargaining with employers, negotiating on behalf of the membership to create an agreement called a collective bargaining agreement, or CBA. In addition to laws, regulations, and court decisions, the CBA governs the terms and conditions of the workers’ employment. A CBA may be used to define the conditions under which a worker could be terminated for cause, require additional procedures for RIF actions, or condition or prohibit the use of automated decision-making systems.  

What’s the problem with algorithms making employment decisions?

Private companies have been using automated decision-making tools in employment decisions for some time now. The creep of worker data collection, surveillance, rating systems, and automated decision-making is called “algorithmic management.” DOGE’s attempts to use an LLM to cull “unnecessary” workers is a form of algorithmic management and automated decision-making.  

Automated decision-making is touted by proponents as a valuable tool for improving efficiency and reducing risks of mismanagement and discrimination. In reality, using these tools come with several core concerns: bias and discrimination; worker surveillance and control; and inadequate transparency and accountability. Using these tools does not absolve the government from its responsibility to follow the law and refrain from violating worker’s rights.   

Automated decision-making tools have significant bias and accuracy issues baked into the tools due in part to the data sources they are built on. Data sources may include limited, incomplete, inaccurate, or illegally used data. Further, automated decision-making tools often repeat and worsen bias because they use historic, inequitable data to produce a decision and lack the ability to understand and adapt for context outside of raw data (for example, a system trained on historic hiring data in engineering would conclude that white men are the best candidates since they have been hired most frequently, ignoring that women and other races were actively excluded from the industry for decades). For these reasons, the government’s reliance on automated decision-making tools to make termination decisions is more likely to lead to discriminatory or arbitrary decisions, potentially in violation of Title VII and other protections listed above. 

Automated decision-making systems feed off massive amounts of information, including information from the workers themselves. Individual worker data could include a worker’s behaviors, communications, associations, personal information, demographic information, and more. The invasive and controlling nature of these tools put additional pressure and stress on a worker while eroding their security, stability, and rights. The current administration has already gone fishing for information on individual employees through improper access of employment records at other agencies. While it is unclear whether the data gathered thus far is being used to build an automated decision-making tool, the uncertainty alone is enough to destabilize and exert control over federal workers. 

Federal employers using automated decision-making tools sharply reduces transparency for workers and their representatives. There is often no insight into how the tool works, what data it is being fed, or how it is weighing different data in its analysis. The logic behind a given decision is not accessible to the worker and, in the government context, it is near impossible to know how or whether the tool is adhering to the statutory and regulatory requirements a federal employment tool would need to follow.  

The opacity and secrecy of automated decision-making tools disrupts the accountability of the government. It is difficult to challenge a decision as illegal, improper, or discriminatory if you cannot access the tool, its information, or any outside evaluations of the system, like audits or validation studies. This lack of transparency and accountability is compounded by DOGE’s equally opaque structure. Its members, decision-making procedures, security clearances, and authority are all in question. In the case of AutoRIF or other automated decision-making tools, this means that unknown actors with unknown levels of governmental expertise are creating tools that will make life-altering decisions on a massive scale with little to no oversight.  

These same issues also erode the federal worker’s rights to procedural due process and appeals. Fair and effective advance notice to the worker should inform the worker that the government used an automated tool to make the termination decision so that the worker can challenge the system. If an automated decision-making tool has been used to execute RIF actions so far, the notices do not make it clear. However, if the worker is not informed of the tool’s use, they cannot adequately exercise their right to respond to the termination or challenge the termination with an appeal. Diminishing a worker’s due process rights further shrouds the government’s activities in secrecy and shelters it from accountability. 

Where do we go from here?

The Administration is attempting to cripple critical public services and weaken checks on the Executive Branch. Automated decision-making tools would allow them to decimate the federal workforce more quickly and make it harder for affected workers to challenge decisions or exercise their rights by obscuring the decision-making process and decreasing insight into the tools. Employing these tools also increase the chances that the government’s decision will be based on illegal grounds. We must protect federal workers from these harmful tools. 

Labor unions, federal workers, and other advocates are fighting to protect federal workers.  Eligible federal workers can act by joining their union to take advantage of the increased job security and just representation. That said, history has shown us that unions benefit all workers. 

Legislators must also move to protect our federal workforce. Existing administrative agencies are not enough to combat the harmful effects of automated decision-making. This is especially true under an Administration which has systematically hobbled labor agencies and deleted prior worker-protective guidance on AI. Congress must act swiftly to stop mass layoffs and ensure the dignity of federal workers. Given automated decision-making tools’ enormous capacity for harm, EPIC urges Congress to prohibit its use in employment decisions without requiring rigorous external testing and auditing, robust notices and disclosure, and human decision review. If the government cannot or will not effectively mitigate the risks of using automated decision-making technology, it should not use it at all.