Major model providers are fighting for the affections of the US government.

In recent weeks, both OpenAI and Anthropic have offered access to their models to government agencies practically for free, with OpenAI proposing to provide ChatGPT Enterprise to the executive branch for $1 a year and Anthropic matching that price to supply its Claude models to all three branches of government.

As these organizations seek to weave their way into government agencies’ tech stacks, however, whether their models can safely handle sensitive private data at scale remains uncertain.

“By now, model providers are well aware of security being a big impediment to adoption,” said Arti Raman, founder and CEO of AI data security company Portal26.

Though AI models still have their kinks, the bigger security problem isn’t the tech itself but the people who are using it, she said. “The bigger risks are on the human side.”

A variety of vulnerabilities have already been demonstrated:

• According to IBM’s recent Cost of a Data Breach report, 97% of surveyed organizations that experienced AI-related security incidents reported not having access controls in place.
• Of those incidents, 60% led to compromised data, and 31% led to operational disruption.
• “Who gets access and access control becomes a bigger problem than data leaking from a model that may not be connected to the outside world … Data security risks are from person to person,” said Raman.

And in government agencies, especially those handling large amounts of sensitive information relating to civilians, the risks can be great. The organizations are often the target of threat actors already, and with the Cybersecurity and Infrastructure Security Agency facing persistent cuts under the Trump administration, the layers of protection may become even thinner.

While government workers are often “trained and conditioned to worry about security,” said Raman, the nascent and evolving nature of AI means that training and governance can’t be a static thing.

“Training and education are incredibly important,” said Raman “It can’t be in the form of a manual or something that you do once a year. It has to be done in real time.”

Education may be only part of the solution. The “white space” of AI access and identity control could represent a major opportunity in the market, said Raman. “We need some innovation on complex identity and entitlement … somebody needs to really understand how to connect the dots between what a model was trained on and has access to versus what it is and isn’t allowed to answer.”

Can enterprises strike a balance between AI augmentation and replacement?

If not, the risks may be immense: A recent study published in Nature found that generative AI’s disruption of the labor market could lead to a socioeconomic tipping point. Using Australian data, the analysis found that even a moderate increase in the “AI-capital-to-labor ratio,” which measures AI investment relative to human labor, would double labor underutilization by 2050.

That could lead to a 26% decrease in per capita disposable income, and reduce the consumption index by 21%, according to the report. As similar reports continue to emerge and new AI laws are enacted, organizations may need to consider the alternative: AI worker augmentation.

Tech companies like Oracle with Oracle Growth and IBM with SkillsBuilds have pledged to upskill millions of workers in AI. Meanwhile, companies like Zoom with AI Companion and startups like Humancore aim to meet the demand for AI that empowers workers.

“It’s shortsighted to see AI only as a cost-cutting tool,” said Mike Dolen, CEO of Humancore. While some tasks can and should be automated, the real upside is thoughtful augmentation, Dolen said.

The key is to embed AI into everyday workflows in a way that’s approachable and supportive, said Smita Hashim, chief product officer at Zoom. When rolling out AI augmentation programs, there are a few points enterprises should consider, Hashim said:

• For one, it’s important to define goals on improving productivity and employee experience. “These goals should map back to the pain points identified by employees, and they should be tied to measurable outcomes so progress can be tracked and celebrated,” said Hashim.
• Tools should be approachable, rather than requiring advanced technical skills. The ultimate goal of AI adoption is to remove barriers, not introduce new ones, Hashim explained. AI support should be seamlessly integrated into employees’ daily workflows, reducing context-switching and minimizing the learning curve.
• Leaders should also provide clear guidance to teams on how they can get the most out of AI, Hashim said.

These kinds of programs draw on organizational psychology and leadership science to ensure every recommendation aligns with the team’s dynamics, shared priorities and cultural realities, Dolen said.

Enterprises need to take care to manage workforce re-education and transformation properly, however. Otherwise, companies risk damaging performance with technological misunderstanding, clashes between management and unions and erosion of brand value and trust.

“Continuous feedback loops are vital to identify pain points, fine-tune AI tech, and build trust among employees,” Hashim said.

As big box stores start to give themselves tech makeovers, Walmart may have its sights set on artificial intelligence IP.

The retailer is seeking to patent a system for “supply chain modeling and prediction,” relying on machine learning to keep track of all the moving parts within its sprawling and complex logistics operations. Walmart’s tech seeks to predict and adapt to real-time changes that may impact the entire operation, addressing limitations in available tools.

“Existing simulation tools are all limited to some specific sub-tasks in supply chain operations, are not adaptive to real-time data and lack the ability to capture the dynamic nature of supply chain operations,” Walmart said in the filing.

When Walmart’s model receives a query, such as “what happens if a distribution center is shut down for two days,” the model will use historical data to build a graph representing the different factors within the supply chain that may be impacted, including suppliers, warehouses and transportation routes.

The model, taking into account the interdependencies of the operation, will use that to generate predictions like inventory impacts, delivery time changes, bottlenecks and costs. This could allow Walmart to consider the domino effects of different incidents from warehouse to delivery to store.

This isn’t the first time that Walmart has displayed an interest in AI. The company has spent the past few years bolstering its tech portfolio, with patent applications including recommendation engines, AI analytics and churn prediction. And the retailer’s tech overhaul is one indicator of a broader trend: Traditionally brick-and-mortar institutions, both in retail and food service, have started to work towards adding AI into their tech stacks.

For Walmart in particular, the shift makes sense. Its biggest competitor has long been Amazon. And while it’d be hard pressed to compete with the tech giant’s AI prowess, developing and adopting its own models – and claiming IP as it does so – couldn’t hurt.