Unity Workers Question Company Ethics As It Expands From Video Games to War

Unity, the popular game making tool, has long been used in industries outside of video games. Unity's site celebrates this versatility, saying its technology offers "incredible possibilities" for film, engineering, architecture, automotive, transportation, and more. 

Unity's site also proudly lists its technology being used by the government and military, but internal Unity documents obtained by Waypoint show the company is struggling to explain why its employees, who supposedly signed up to create tools that empower game makers, are now directly or inadvertently developing technologies for militaries with the stated objectives of fighting wars. 

According to three sources Waypoint talked to, some Unity employees might develop technology that ends up with military clients without even realizing that's how their work would be used. These sources, a mixture of current and former Unity employees, were granted anonymity to avoid reprisal.

According to these sources, the part of Unity that's pursuing government and military contracts is sometimes nicknamed "GovTech," which the company described in a public presentation from March 2021 as intended to "develop technologies across our products that helps the government adapt AI and ML [machine learning]"

One internal memo, titled "GovTech Projects - Communication Protocol," which was shared with Waypoint but not distributed widely within Unity itself, outlined how the company should talk about these contracts with "internal/external stakeholders," and makes clear Unity understands the delicate line it's walking. 

"We need to be sensitive to the various values & beliefs which people perceive our engagement with the Government, specifically DoD [Department of Defense]," reads the memo, which instructs managers to use the terms "government" or "defense" instead of "military."

The memo is listed as a draft, isn't dated, and Waypoint's sources were unaware of a final version meant for wider distribution. It includes a list of "Do's" and "Don'ts" for Unity employees when talking about GovTech projects. Under "Do's," the memo tells employees to point out that Unity is using AI to improve how the DoD runs simulations and trainings, and that "Nothing we are doing will be used in live warfighting." The memo also instructs employees to highlight that Unity's "current projects provide a service and or solution to DoD companies and we are not taking the lead on any single project."

Under "Don'ts," the memo instructs employees not to "discuss any projects that involves the use of simulated or virtual weapons or training to harm another person."

A postscript on the memo also implies that Unity's work with the military is justified by the DoD's own ethical principles. "PS: It's important to note that DoD itself has published a set of AI principles that it has pledged to adhere to as it adopts AI at scale," the memo says, and links to a 2020 press release by the DoD that outlined its own "series of ethical principles" related to the research and use of military artificial intelligence. The memo highlights the following quote from the DoD press release: "The principles address these new challenges and ensure the responsible use of AI by the department."

Unity started as a Mac-exclusive game making tool in 2005, but in 2021, it's become one of the dominant game development tools in the industry, powering everything from indie hits like Hollow Knight to big budget games like Hearthstone. Much like Unreal Engine, Unity offers game developers a variety of tools to create games. Unity is known mostly for tools that allow developers to create 3D environments and models, but developers also use it to script what happens in a game, create the software that dictates how non-player characters behave, and more. If you're regularly playing games these days, there's a good chance you've encountered the company's white cube logo on a game's opening splash screen.   

The company went public in 2020, and is currently valued at $33 billion.

Most Unity employees do not have immediate access to every detail about a project the company is working on. Employees can request that information from a manager in some cases, but it's rarely easy. Part of the problem, sources told us, was that not all Unity employees knew exactly what Unity was doing for the military, and if the projects that they were working on could end up supporting Unity's work for the military without them realizing.

"Most Unity AI work empowers other government projects, so in this way it can be difficult to gauge one's contribution to government projects," said one source.

According to one source, a Unity engineer could be working on an AI tool without a specific application in mind and have no way of knowing if another part of the company would then use that tool for a contract with the military.

"It didn't seem very clear through company training or anything like that there even were concerns, really, in what we were doing," the source said. 

"It should be very clear when people are stepping into the military initiative part of Unity," said one source.

According to a document Unity employees used to collect answers from managers to ethical questions, and that was shared with Waypoint, an unnamed employee noted they'd been tasked with working on what had been pitched as "a placement randomization scheme for a government simulation project." In reality, according to that unnamed employee, it was "simulating explosion debris on virtual runways" for the United States military. The same employee noted how Unity frequently used "government" to describe military projects, a practice outlined in the memo Waypoint obtained.

"I don’t think most employees were warned that they would need philosophy degrees before accepting a position at Unity," reads a different comment from another employee. "Many of my fellow employees are very talented engineers or artists, but I don’t think that experience necessarily prepares us to be successful in passively identifying the complex repercussions of advancing certain emerging technologies."

In response to an inquiry about this story, Unity told Waypoint that one of the company's goals is to "have applications outside of gaming." The company did not respond to a list of specific questions, which included inquiries about whether Unity employees had ever knowingly or unknowingly participated in the creation of a weapon of war and how much the company makes off its military contracts, and instead passed along a long statement included in full at the end of this article.

In that statement, Unity said use of its technology by governments and the U.S. military "is not new" and pointed to its current and past contracts with the U.S. Air Force, U.S. Army, and defense contractors, such as Lockheed Martin and Boeing. Pressed for specifics on those contracts, Unity said it "can't share specific details" and said it has a "thorough review process" so the company doesn't "knowingly violate our principles or values."

A key principle, according to Unity, is that its work for the government "does not directly involve the loss of life, harm of the planet, or a person's right to equity and inclusion." Citing confidentiality clauses, the company could not point to specifics, but brought up its internal Sales Ethics Advisory Council (SEAC) "made up of interested Unity employees from diverse backgrounds, geographics, and parts of the company." Unity said the SEAC has passed along recommendations that have resulted in declining contracts "that do not align with the principles outlined above."

"We believe in the freedom of expression, and as the Unity engine is a tool, a tool that can be downloaded and used by most anyone, it’s impossible to truly police or censor all uses of it," Unity said.

After Waypoint reached out for comment, Unity CEO John Riccitiello released a similar statement internally to employees over a company-wide Slack on Thursday, a copy of which was shared with Waypoint.

"We expect a story from a VICE journalist to talk about Unity’s work with the government and military," Riccitiello wrote in the Slack message. "The contracts for these engagements are very restrictive - similar to many of our partnership deals including some of the work we do with gaming companies and in verticals such as retail - so while we can’t share specific details, we can say that we have a thorough review process, and we have not nor will we support programs where we knowingly violate our principles or values."

After Unity went public in 2020, it pursued expansion into other sectors, which has in recent years included artificial intelligence contracts with the U.S. government, specifically the military. 

"At Unity, we fundamentally believe the world is a better place with more creators in it," wrote the company in a blog post about its "principles for ethical AI" in 2018. "Artificial Intelligence (AI) is a key component of that mission."

That same blog laid out a few specific goals for the company's initiatives, including one called "be accountable," wherein Unity would "consider the potential negative consequences of the AI tools we build" and "anticipate what might cause potential direct or indirect harm." Another point, called "be fair," outlined how Unity would not "knowingly develop AI tools and experiences that interfere with normal, functioning democratic systems of government."

Unity's work with the government isn't a secret—there's a giant promotion for this collaboration on Unity's website—but its existence has created tension, thanks to the increasingly contentious merging of the military and AI.

"This has been largely received poorly internally for those that know about it," said one source.

Waypoint uncovered three government contracts with Unity in 2020 alone, including a $428,000 contract in June 2020 with the U.S. Air Force for a "modeling & simulation prototype" that would function as a "multi-domain operations advanced battle management family of systems." The Air Force signed a new $220,583 contract with Unity with the same description in October 2020, as well. The third contract for $23,500 was for a Unity Pro subscription for the U.S Army. The online records also showed the Navy paying for Unity licenses in the past, including in 2012 for $4,800.

An internal presentation titled "Unity AI for government" breaks up the company's strategy in this space into three stages: Crawl (adoption), Walk (recurring revenue), and Run (self service). In the first "adoption" phase, the presentation says Unity aims to enable developers by "hiring thought leaders and subject matter experts," building a "cadre of govt and systems integrator evangelists," and raising awareness by promoting "purpose-built demos" and educating "the government community." 

The presentation then goes on to explain the "Government AI opportunity" by listing how much difference governments spend on AI, showing that U.S. defense and intelligence spent $1.4 billion in 2020, the most on the list and a 43 percent increase over 2019. It also lists Unity's "key products" in this space, including "Robotics," which is detailed as "Prototype robotics systems, test robotic applications, train robotics algorithms and operate robots with simulation." Another key product is "Computer Vision," and a different slide in the presentation mentions a case study for "Dangerous Object Detection," and shows a computer vision model to "detect bats, guns, and knives." 

One example of how the Department of Defense uses Unity is RIDE (Rapid Integration & Development Environment), a simulation creation platform made at the University of Southern California that is, according to the university's own website, "in direct support of Department of Defense-funded research" and sponsored by a contract with the U.S. Army Research Laboratory. While it's built to support other game engines, it's originally made in Unity.

A video promoting RIDE mentions incorporating "cutting-edge artificial intelligence research," while showcasing simulations involving piloting a tank, a group of soldiers firing weapons, and more. A June 2021 article profiling RIDE in SIGNAL Magazine, a publication focused on topics like information security and "homeland security," noted RIDE would eventually include "AI to enhance the behaviors of simulated characters, including enemy forces," because "military simulations need to accurately depict behaviors of civilian populations."

Unity also conducts private conferences pushing Unity's applications for the military. This past March, Unity hosted a talk about "applied research & prototyping in defense applications" specifically meant to showcase "Unity's latest technological developments for applications in the government sector." The guests included developers on the RIDE project.

"We are here and committed to being the U.S. government's real-time 3D platform of choice," said Unity head of government solutions John Cunningham during the talk. "Our government team is acutely focused on identifying how we can leverage [new Unity] technologies for government and defense and even to develop new products if necessary."

All three sources told Waypoint the current debates at Unity about government contracts can be traced back to the company's decision to partner with the oil and gas company Schlumberger in 2019. The Guardian once described Schlumberber as "the oil world's most secretive operator" and "the biggest company you've never heard of." The company operates all over the world and was fined $155 million in 2015 for violating sanctions in both Iran and Sudan.

As part of the 2019 partnership, Unity helped Schlumberger move its data into the cloud. But this deal sparked an internal backlash at Unity and arrived simultaneously with increased scrutiny from Unity employees about the contracts the GovTech department was prioritizing. 

"There was an increasing trend for the AI org to focus on GovTech professional services projects," said one source. "This felt uncomfortably close to the military."

At an all-hands meeting this year, according to one source, an employee submitted an anonymous question asking if there were any GovTech partnerships at Unity that didn't involve the Department of Defense. The company could only name one.

The backlash to the Schlumberger deal and the increased ethical concerns caused Unity to form the Sales Ethics Advisory Council (SEAC), according to Waypoint's sources. Unity said SEAC was the result of "a series of company-wide roundtables with hundreds of Unity employees a couple of years ago."

In theory, the SEAC would vet new projects and hand recommendations to executives, including CEO John Riccitiello. Riccitiello was president and COO of Electronic Arts from 1997 to 2004. He left, then returned as CEO from 2007 to 2013, before ultimately becoming Unity's CEO in 2014.

Similar ethics groups have become trendy in Silicon Valley. In March 2019, Google announced a similar "council" for its own AI programs, to "guide the responsible development and use of AI in our research and products." Immediately, one person Google initially announced to be on that council said they would decline the invitation. A petition resulted in another person resigning. Less than a month later, the whole initiative was disbanded. 

One of the chief criticisms of the council, according to Waypoint's sources, is the lack of transparency about how the council operates, when and why it decides to review a potentially problematic contract, and how it comes to a decision. Frequently, employees are given limited insight on a project and may not have the whole picture on how a given technology might be used in the future, or the possible negative consequences. They're relying on the council doing the ethical work, but the reviews are not made public to individuals outside of the council, making it unclear what ethical lines Unity will or won't draw. This, combined with a lack of transparency and understanding about how Unity technology is used by the military, is what's lead a number of Unity employees to demand more answers from the company.

The person who makes the final call on projects, including those with the military, according to the document Unity employees used to collect answers from managers to ethical questions, is Riccitiello, and that "the assumption is that he will act upon the [council] recommendation." But the council recommendation is not required, which underscores the existing criticism of the arrangement: that the council lacks real teeth. 

In a statement to Waypoint, Unity said that SEAC recommendations have resulted in the company rejecting "deals that do not align with the principles [of Unity]." 

In April of this year, an internal AMA (ask me anything) session was held on Zoom among the company's AI group, which allowed a broader selection of employees to ask questions about projects the company was taking on and how it intersected with its stated AI responsibility principles and the role of the SEAC council in the company.

"It was not quite the disaster that I thought it was going to be," said one source who attended, "but it was great in that a lot of people had questions and they [Unity] had very few answers."

"I came to Unity explicitly because I naively believed their marketing around 'empowering creators' and 'making the world a better place' or whatever," said one source. "I got into AI with the expectation that I'd be building technology 'for the greater good' or some nonsense like that. You learn pretty quick, though, that cozying up to warfare profiteers is the fastest way to make money pretty much universally in the tech industry."

"Whether or not I'm working directly for the government team, I'm empowering the products they're selling," said another source. "Do you want to use your tools to catch bad guys? Maybe we shouldn't be in the business of defining who bad guys are."

Riccitiello’s Slack message anticipating this article provoked an enormous response from Unity's more than 4,000 employees, most of whom do not participate in or around Unity's government work. One employee told Waypoint there are more than 100 responses to Riccitiello’s message, a discussion still ongoing, with hundreds more responding with emojis. 

"The reactions are mixed," said the same source. "The largest group is angry to be finding out we're working with the military at all. There's a group that is confused or upset, but is withholding full judgement until they read the article. There's a group repeating some version of the 'slippery slope' argument over and over, and then there's a small contingent of men (mostly executives and upper management) who think we should all just Support 'Our' Troops and anything short of actually dropping the bombs ourselves is totally fine."

The response was enough to prompt a follow-up by Riccitiello, promising to open Unity's next town hall meeting, scheduled for Tuesday, to start with a discussion about the topic.

"I would not want Unity to do anything that violates our principles," said Riccitiello. "This is super-important to me. I realize that each of us would have their own individual take on where to draw the line on what works and doesn’t."

Riccitiello praised the company's internal watchdog council, SEAC, though underscored it functions in an advisory role—it does not accept or reject deals on Unity's behalf. He noted SEAC is "relatively new" and Unity doesn't "have well practiced and enforced policies."  

"Some of what we’re reading is 100% accurate," he said. "Some not. Some a perspective based on partial information. I don’t see Slack as a good place to deal with these issues."

Riccitiello promised "full transparency" in Tuesday's town hall meeting.

Unity's full response to this story is below:

Additional reporting by Joseph Cox.