Generative Artificial Intelligence is the AI produced primarily by Large Language Models, or LLMs. These LLMs reference the information they consumed from the Internet and other sources during their training. They additionally search for and incorporate newer information provided by their trainers at the AI companies. LLMs recognize patterns and use them to create new content based on what they have learned. Generative Artificial Intelligence (GenAI) agents can achieve objectives based on goals without human intervention to provide task-level instructions needed by other automated systems.   

Transportation has previously embraced technology and in doing so made important progress. Over the years transportation operators used automation based on a variety of methodologies that include strict parameters and business rules, machine learning, and non-generative artificial intelligence. The systems these technologies support include autopilots, autonomous and assisted driving, decision-support tools, and automated train operation. Computer vision technologies have been used for license plate readers, bus lane enforcement, red light and speed cameras, and transit signal priority, among other applications. Travel apps have used legacy AI to calculate best, fastest, and least expensive routes and reliable predictions of arrival times. Electronic toll collection devices, cargo port cranes, and Unmanned Aerial Vehicles and Vertical Take-off and Landing vehicles all use forms of artificial intelligence.  

GenAI’s capability to create new content, make independent decisions, and act without human intervention may advance the capabilities of transportation automation to increase safety, provide greater levels of accuracy, improve mobility, and help create and deliver new services. GenAI chatbots, coders, and agents have the potential to improve the productivity of transportation knowledge workers including planners and engineers to provide more value to taxpayers and to customers. Transportation staff should learn about GenAI and how it can help them and their operations. However, operators first should consider the following questions before deploying GenAI, especially in mission critical and life safety systems: 

1. Should government and transportation operators use GenAI knowing it lacks important safety controls that prevent it from causing harms to people, society, and the planet, including the following examples?  
   1. Violating privacy, personal agency and security. Prior to GenAI, surveillance technologies including roadway sensors and cameras and even cell phone tower connection information could reveal people’s travel movements and their home and work locations. GenAI allows this information to be more easily combined with data from technology such as shared mobility devices including bikes and scooters that may obtain personal data without explicit permission. GenAI can use this exposed personal data to create more comprehensive profiles of personal mobility that could additionally threaten personal security and agency to move freely.  
   2. Undermining democracy, authenticity, and legitimate authority. GenAI can be used to create realistic but fake telephone communications, movies, and still images to interfere in elections. It can impersonate public safety officials in fake emergency communications obstructing national, regional, and local transportation operations. Authoritative-looking fake messaging can be used to create panic or direct passengers into harm’s way during an emergency event.
   3. Weakening cybersecurity and degrading commercial and national security.  GenAI currently provides a demonstrated advantage to cybersecurity adversaries over the private and public institutions and their staffs who safeguard electronic commerce, network communications including the Internet, and protect the nation. Even prior to GenAI’s wide availability, nation-state adversaries breached critical infrastructure systems including those in transportation. They have listened in to our communications and have accessed critical infrastructure control systems. GenAI’s increased capabilities have been demonstrated by GenAI agents from OpenAI attacking and breaching in July 2026 the open-source AI repository managed by Hugging Face (yes, that’s its real name). 
   4. Causing job displacement and losses. This harm may be felt throughout the public and private sectors if companies use GenAI to replace workers or drive down their wages and not enhance and support the workforce. The transportation sector relies on many private sector firms for operations, design, engineering, and equipment. Use of GenAI by these firms may cause job displacement even if transportation agencies do not directly use GenAI or displace workers. 
   5. Widening existing wealth and income gaps. Job loss and wages driven lower by the use of GenAI would redirect economic benefits from workers to large GenAI firm owners and investors. Medium and small company owners and investors that develop their own LLMs as well as apply large company LLM technology to the transportation sector may also receive the funds that once paid workers. They would receive these funds to pay for the use of GenAI agents, coders, and chatbots. These changes in the flows of economic benefits may worsen current income and wealth concentrations that already accrue 20%-50% of income and wealth to the wealthiest 1% of the US population. 
   6. Decreasing national financial stability. Financial analysts have identified a number of factors illustrating that GenAI presents large-scale financial risk to the US economy and leads many analysts to believe we’re experiencing a financial AI bubble.  First, seven companies account for nearly 30% of the market capitalization of the entire US stock market. All seven of those companies spend billions of dollars yearly on AI. Analysts believe five of the seven spend over $100 billion yearly on GenAI. Decreases in performance or reductions in investments by just a couple of those firms may lower performance of the entire market due to their dominant influence. Second, many financial analysts describe the AI economy as circular. Investments from cloud and data center companies such as from Amazon or Google to AI companies such as to OpenAI and Anthropic wind up back at Amazon or Google as payment for cloud services from the AI companies.  Similar circular activity occurs between hardware manufacturers such as NVIDIA when they invest in cloud and data center companies.  The same funds make the rounds to multiple companies. Like in a game of musical chairs, an interruption in the flow among those companies due to a market or company problem may leave some companies without a seat and threaten their survival which may spread to the larger economy due to their dominance in the market. 
   7. Increasing global warming. GenAI data centers so big one is named Colossus, have an insatiable need for power which the current electricity grid cannot satisfy. The Colossus construction modeled off-grid fossil fuel powered generators to create electricity for GenAI data centers. They add to the carbon load expelled into the atmosphere, increasing everyone’s temperatures. 
2. Is GenAI development a good expenditure of government resources including cash expenditures and use of government staff time? How do we evaluate the return on investment (ROI) when employing these and other resources to deploy GenAI in the public sector? Government agencies have no Profit & Loss statements to help evaluate whether spending on GenAI will produce a positive ROI, and quantifying public benefits for ROI calculations can be complicated and require years of data.  
3. What impact will individual GenAI customers experience if the AI bubble bursts, the stock market rapidly and significantly declines, and some GenAI companies close? Currently LLM companies are not profitable. Their stability depends on revenue from private investors and cloud company investments, not from fees paid by their GenAI-user customers. In other words, GenAI customers benefit from below-cost pricing with investors paying the difference. What happens if organizations that rely on below-cost pricing to achieve a positive GenAI ROI receive large price increases to continue using the technology after a market or company failure? What happens to organizations that went all-in with a single GenAI company that cannot survive the bubble bursting? Organizations that prepared reserves or designed a multi-vendor strategy may survive that storm. However, after an AI market disruption under-resourced transportation operators and governments that made significant moves to use GenAI technology may have exposure they don’t have funds to cover.  That exposure can lead to service degradation or interruption. 

Transportation has lots of experience with early forms of artificial intelligence which helped the sector offer safer and better services. Transportation operators should hurry to become familiar with and to safely experiment with GenAI. Perhaps they should begin to use it in back-office and administrative applications, provided its risks in those environments get sufficiently addressed and it offers better service than current systems. But, operators should wait to deploy it in transportation operations until we have sufficient regulations in place. Laws should require LLM companies to create guardrails to sufficiently manage its risks to life safety and mission critical systems, and protect people, society, and the planet from potential harms.

Cordell Schachter is Principal with Riverdale Advisors. He formerly served as CIO, US Department of Transportation, and as CTO, New York City DOT.