This is not to be confused with LiDAR, for Light Detection and Ranging, a kind of remote sensing system that uses laser pulses to measure distances from the sensor. The light-specific LiDAR, sometimes also written LIDAR, is a valuable tool for remote sensing and mapping, and features prominently in the awareness tools of self-driving vehicles.

Purdue’s LIDAR, instead, is a kind of architecture for network security. It can adapt to threats, thanks in part to its ability to learn three ways. These include supervised machine learning, where an algorithm looks at unusual features in the system and compares them to known attacks. An unsupervised machine learning component looks through the whole system for anything unusual, not just unusual features that resemble attacks. These two machine-learning components are mediated by a rules-based supervisor.

“One of the fascinating things about LIDAR is that the rule-based learning component really serves as the brain for the operation,” said Aly El Gamal, an assistant professor of electrical and computer engineering in Purdue’s College of Engineering. “That component takes the information from the other two parts and decides the validity of a potential attack and necessary steps to move forward.”

By knowing existing attacks, matching to detected threats, and learning from experience, this LIDAR system can potentially offer a long-term solution based on how the machines themselves become more capable over time.

Aiding the security approach, said the researchers, is the use of a “novel curiosity-driven honeypot,” which can like a carnivorous pitcher plant lure attackers and then trap them where they will do no harm. Once attackers are trapped, it is possible the learning algorithm can incorporate new information about the threat, and adapt to prevent future attacks making it through.

The research team behind this LIDAR approach is looking to patent the technology for commercialization. In the process, they may also want to settle on a less-confusing moniker. Otherwise, we may stumble into a future where users securing a network of LiDAR sensors with LIDAR have to enact an entire “Who’s on First?” routine every time they update their cybersecurity.

“The private sector wants to keep the lights on just as much as the public sector does,” said Sean Plankey, principal deputy assistant secretary in the Department of Energy’s Office of Cybersecurity, Energy Security, and Emergency Response.

To that end, Plankey pointed to trainings sponsored and coordinated by the Department of Energy that bring together industry and government to model crisis simulations. Some of these are modeled on the “Black Energy” attack by Russia on Ukraine’s power grid, and show the importance of integrating federal emergency response with private-sector handling of critical infrastructure.

Convincing stakeholders within government of the importance of cybersecurity is but one part of the job. Another is making sure that skilled people can be recruited, brought on board, and then kept engaged in government careers so that they are available when needed.

Looking at the available labor statistics for employment in the industry, New America Cybersecurity Policy Fellow Laura Bate found that, while the private sector paid more for the same degrees than government, it amounted to only about a 24 percent increase. While that’s a significant gap, it’s a much smaller and more manageable disparity than popular perception, which holds that the private sector pays orders of magnitude more than the government for the same skill set.

“We don’t have to start people at step one,” said Venice Goodwine, chief information security office for the Department of Agriculture. “There are opportunities for negotiation when entering government.”

Simply taking advantage of the tools on hand, and knowing the full extent to which those tools can be used to build the workforce at all levels, is one key way to mitigate that disparity. Another is emphasizing the full range of the mission at places like USDA, where cybersecurity can matter for everything from crop integrity to the safety of people fighting forest fires.

And while the perception is that technology companies are what’s driving cybersecurity salaries, Mantech President Rick Wagner highlighted that the banking industry is now a huge competitor for that same supply of workers. Government has to offer something beyond just the dollars to be competitive for those workers.

“If I want to keep the young, bright-eyed, bushy-tailed individuals, I need to make an environment where they can be creative,” said Goodwine. To than end, Goodwine emphasized reskilling to cyber from within government, and finding ways to pay for cyber workers to advance their own knowledge, so that they can stay engaged and keep doing the job.

The necessity and difficulty of clearances was a recurring theme for the panel.

“Clearances, and the delay to get people from conditional offer to into the job bigger deal than numbers suggest,” said Bate. “Anecdotally, I can say it’s a huge part of why people walk away; as a researcher that we can’t quantify it drives me crazy.”

Goodwin spoke to USDA having a fast turnaround of between 30 days and six weeks for interim clearances as one way to bring people in who might be turned away by the process. On the civilian side, the presence of cybersecurity jobs that don’t require clearances is also helpful to bring people into government.

“When having people break down and cry during polygraph is part of the normal hiring process, you’re not starting the relationship with an employer in the right way,” said Bate.

For Plankey, making clearances sticky for people, allowing them to have clearances sponsored while they work in the commercial sector, would make it easier for them to return to government service, even if it isn’t the same part of government they worked at previously. And there are more unconventional avenues to get people into cybersecurity roles.

“The NDAA authorized direct hiring into officer ranks for cyber expertise,” said Plankey. “The authority is out there; experience using it is not.”

At stake is everything from the basics of how civilians in the Pentagon make PowerPoint presentations to how the whole Department of Defense moves to 5G.

While the implications, for now, center on top-level manufacturers, the possibility exists that in an effort to reach perfect security, the component supply chain is radically reshifted. That the modern world is bound together by global supply chains is not news. An attempt to fully decouple those supply chains will invariably become news, depending on exactly how deep into hardware the quest for Made-in-America goes.

On Nov. 6, 2019, Sen. Marco Rubio sent an open letter to the Pentagon, asking the Department of Defense to clarify what steps it had taken to phase out Chinese-made technology. Rubio’s letter highlights the findings of a report by government contractor Forescout, which found 3,500 devices “from telecom giants Huawei and ZTE, as well as surveillance camera-makers Dahua and Hikvision, on U.S. government systems a month before the ban was to take effect.”

It is the connectivity of these devices, at least as much as the nation of origin, that makes it a security risk. A camera that can connect to an online security network is one that could potentially share what it sees remotely. The ban on such devices, which went into effect in August 2019, specifically concerns technology produced by companies with links to China’s government.

The interim rule, already written into the 2019 National Defense Authorization Act, followed the July 2019 publication of Department of Defense Inspector General report that found the Pentagon spent over $32 million on risky technology in 2018. This spending included everything from Lexmark printers to GoPro cameras to Lenovo computers, all major commercial off-the-shelf products and all, according to the report, with cyber vulnerabilities.

The printer, camera and computer markets all have companies in them with headquarters in the United States, so there’s always a way for the Pentagon to buy domestic instead. But if concerns spread to fear of compromise from not just the top-level of the company, but harm in the manufacturing process, then the interconnectedness of international manufacturing may show where a company is headquartered to be far less relevant for security than where its product is built.

In October 2018, server motherboard company Supermicro become the subject of one such fear. While Supermicro is headquartered in Silicon Valley, a report published in Bloomberg suggested that its motherboards were assembled in China and some were fitted with a tiny microchip that allowed China’s government to compromise any network connected to a Supermicro server. While members of Congress have pointed to the case of Supermicro as an example of the threat they are trying to prevent, subsequent investigations into the motherboards were unable to prove the claims made.

In 2019, Supermicro shifted its manufacturing outside of China. It was a move that appeared driven as much by business concerns as by the new considerations of security risks, even if unproven. It was an uncoupling, perhaps of necessity.

In the quadcopter and commercial drone market, where no U.S.-made product matches the price point desired by civilian government agencies, those agencies have worked with manufacturers abroad to develop product that matched government security needs. At least, until unspecified security concerns grounded the fleet of drones.

While there are quadcopters made to the security specifications of the U.S. military, they are not at the same price point, and still feature some components made abroad. Any decoupling of supply chains would necessitate changes in domestic production as well.

There are alternatives to a nation-of-origin based security approach for incorporating commercial technology into government and civilian use. One such approach focuses instead on testable, verifiable standards.

The Center for New American Security published a thorough report Nov. 7 on how the United States can adapt to the challenges of 5G competition. Among other policies, the report recommends the United States take an active role in developing international standards for 5G, especially with an emphasis on “the requirements of cybersecurity and methods to mitigate the risks of disruption must remain a core consideration, including the continued promotion of security standards for internet of things devices.”

The CNAS report goes on to outline a holistic approach for the Pentagon to handle both 5G and the security challenges it poses. A nonzero part of that approach involves investing in domestic industry to develop the capability, and in waiting until that technology is available.

“At present, there are no American companies in 5G that can compete directly with Huawei in the radio access network. There are a small number of viable alternatives among current companies,” suggests the report. “The U.S. government should explore opportunities to diversify and rebalance its existing dependencies in supply chains and vendors. It is important to mitigate dangerous dependencies or the vulnerability that any single player or source of equipment could be compromised by a potential adversary.”

In the meantime, as the Pentagon seeks to decouple its supply chains from foreign production, it will have to wait for industry to catch up to the new demand signal.

The announcement came Oct. 30 and was opaque as to what, exactly, the new threat was. It’s a new frustration in what had been the U.S. government’s smoothest relationship with China-base drone giant DJI.

“We are aware the Department of Interior has decided to ground its entire drone program and are disappointed to learn of this development,” DJI said in a statement. “We will continue to support the Department of Interior and provide assistance as it reviews its drone fleet so the agency can quickly resume the use of drones to help federal workers conduct vital operations.”

The use of Chinese-made drones by Interior, specifically those by drone giant DJI, had previously been heralded by Interior and DJI as a way to overcome security concerns while still getting the affordability of commercial drones. DJI noted that the company spent 15 months working with the officials at the Department of the Interior, independent cybersecurity professionals and experts at NASA to bring the security sought by the government to the drones it purchased.

As a result of this collaboration, DJI and Interior announced the launch of DJI’s “Government Edition,” for DJI drones, specifications that included custom software, firmware, and hardware for DJI-made Matrice 600 Pro and Mavic Pro drones selected for use by the interior department.

In its announcement responsing to a previous Interior evaluation of Government Edition drones, DJI said “We’re excited to see the report validate how DJI products can meet the stringent data needs of high-security customers like government agencies and critical infrastructure operators. That’s important, because DOI found DJI drones were far better than those from other companies for their missions.”

Interior’s own assessment was more sterile in its language but supported the same conclusion, noting that the final tested version of DJI’s government edition “provided a reasonable mitigation for known data management assurance vulnerabilities of the stock Matrice 600 Pro and Mavic Pro UAS.”

DJI has such a place in the hobbyist and commercial drone market that, at multiple price points and capabilities, there are few viable alternatives, and despite an explicit interest on behalf of the US government that it buy American, there is no such American alternative on hand.

In 2018, Interior tried to buy quadcopters from 3DR, an American drone company that has since transitioned to making drone software. Without an American quadcopter to buy at the same price point, government officials have picked up DJI drones instead.

Government Edition marked a deliberate effort to match the present need of the Interior Department for a low-cost quad- and multicopter drone with the security concerns that come with buying flying robots made abroad. It was a long overdue acknowledgement that, while DJI did not make drones with the explicit purpose of selling those drones to government, people in the government were buying those drones.

Military buyers have bought and used so many DJI drones that the Pentagon has multiple times issued moratoriums on the use of foreign-made drones until security reviews can be completed. Congress has introduced legislation specifically banning the purchase of drones made by “strategic competitors,” which mostly just means China.

In light of that, the DJI and DOI collaboration of Government Edition for drones was supposed to provide an alternative on the civilian side of government. (DJI is explicit that its products are not designed nor marketed for military use.)

The new flight moratorium on DJI drone use, and the security review, could put an end to that detente between the drone maker and government customers. If it does so, it will leave unfilled the void in price, performance, and scalability that DOI itself identified when evaluating the drone market.

Modern life is millennia removed from proto-ethical debates over the dangers of collaborating on hunting technology with people who may someday turn it to violence, but dual-use tools are at the center of a major inter- and intra-national debate. The U.S.-China Economic and Security Review Commission held a hearing June 7 about China and technology, specifically the ways in which developments in the civilian sector could be exploited and weaponized by China’s military.

“China has been hyped as an AI superpower poised to overtake the U.S. in the strategic technology domain of AI,” said Jeffrey Ding, China lead for the Center for the Governance of AI, Future of Humanity Institute, University of Oxford; D.Phil. Candidate, University of Oxford.

“Much of the research supporting this claim suffers from the ‘AI abstraction problem,’” Ding continued, “the concept of AI, which encompasses anything from fuzzy mathematics to drone swarms, becomes so slippery that it is no longer analytically coherent or useful.”

Into the fuzziness of that debate around AI, the commissioners probed and the witnesses testified, occasionally going into the weeds of specific data but largely moving back to questions of intent and capability. The big questions underlying the hearing were these: can China leapfrog development of AI to create useful military tools, and can the United States marshal an industrial policy to maintain a lead in AI without the same kind of integrated state apparatus that China has? Also, what will China do with AI once it develops it?

The degree to which China can expand on and leapfrog U.S. AI development hinges on everything from the respective countries immigration policies, a willingness of multinational corporations to share research, direct and passive funding, and the extent to which things like counter-intelligence operations can yield results or chill collaboration. It also depends on what purpose, exactly, the AI is being trained to do.

“AI appeals to the [People’s Liberation Army] in part because it has fewer legacy systems,” said Elsa Kania, adjunct senior fellow, Technology and National Security Program at the Center for a New American Security and research fellow at CSET. Building systems for the first time with AI in mind is a different and distinct proposition from adapting existing systems to draw on AI tools.

Looking for a hard metric in the recent past, Ding cited an unpublished manuscript by Jon Schmid, a doctoral candidate at Georgia Tech, that looked at military patents to see what was being developed.

“It’s not a perfect indicator because a lot of military hardware isn’t going to be patented open source,” said Ding, “but a crucial caveat I’ll add to that is a lot of these advanced military systems source components that are found in patents, so it’s a good indicator.”

Looking specifically at the abstracts of patents for military hardware or hardware components that cited autonomous or unmanned, the research found that from 2003 to 2015 the United States had a lead in cumulative military patents seven times that of China.

“We can think of AI as next wave of software improvements,” suggested Helen Toner, director of strategy at Georgetown University’s Center for Security and Emerging Technology. Without a clear inflection point for when AI transitions from capable software to a unique capability, many of the changes will in nature and implementation resemble software upgrades.

AI can be trickier to quantify, because while end-products may make note of features, like autonomy, that are enabled by AI, AI itself is a process done to data and doesn’t always transfer easily or at all to other applications. Also, processes are limited to being only as good at the quality of data that goes into them.

“It’s not quite as simple as ‘data is the new oil,’ but the value of data is really application dependent, so we have seen strength in particular applications where massive data is an advantage,” said Kania. “The nature of AI is a general purpose technology that is enhancing and augmenting weapon systems across a vast array of weapon systems, drive dynamics hard to evaluate, make safety and stability.”

Ultimately, what matters most in AI is less about any specific development, and more about how that AI facilities and enables existing choices people were already prepared to make. Multiple members of the U.S.-China Economic and Security Review Commission asked specifically about the role of AI in China’s repressive policy towards the Uighur people of Xinjiang. That specific concern was cited by Sens. Martin Heinrich, D-N.M., and Rob Portman, R-Ohio, in introducing a bill to enhance U.S. funding of AI research.

“The reason Xinjiang is happening is not AI, it is that the government of China is willing to make it happen,” testified Toner, “and therefore our response, if we respond, should condemn or otherwise sanction that action, of which tech is only a small part.”

The extent to which tech is especially responsible for state-led repression was contested by the panel, with Kania testifying that surveillance technologies are at the heart of repression at Xinjiang.

It is hard for the technology sector to, on its own or through U.S. government encouragement, try to change the behavior of a government. There was somewhat more consensus on policy suggestions for companies and government to end relationships building AI tools used for repression in the name of internal security.

Military implications of AI competition are likely downstream from these tangible concerns over AI built for other purposes. An ecosystem built on top of open-source tools and massive collections of data could be rebuilt if major participants stopped collaborating with each other, but it would likely take a national initiative to coordinate such a break. In the meantime, an active Cold War in AI will remain a near future possibility, rather than an immediate course of action.

Heinrich, D-N.M., wants the bill — cosponsored by Rob Portman, R-Ohio, and Brian Schatz, D-Hawaii — to fund better, more useful and, most importantly, more ethical developments in artificial intelligence throughout government and industry. The danger, central to the AI-IA pitch, is that without a concerted American effort, the future of AI will be decided by forces outside the influence of the United States.

“China’s emergence in the AI space poses a great danger to an ethical global adoption of standards and uses of these technologies,” said Heinrich. “They are not a free market economy that respect civil liberties and privacy the way that our country does. We can already see the danger posed by this in the Chinese use of AI and mass surveillance technologies in their oppression of minority Uighurs in western China. We simply cannot allow this type of unethical use of AI technology to proliferate around the globe or to certainly take root here. That’s what’ll happen unless we step up our efforts here domestically.”

The senator admitted that the United States is unlikely to ever match China in terms of dollar amounts spent on AI, but instead will be able to out-innovate in the AI space by using what resources it does allocate smarter, and more inclusively. Part of this means constructing an AI development pipeline at education institutions, including at least one K-12 school and at least one minority-serving institution. The text of the act notes that it’s important to collaborate “with minority-serving institutions in order to facilitate the sharing of best practices and approaches for increasing and retaining underrepresented populations in the artificial intelligence field.”

Not explicitly stated is the possibility that working with underrepresented populations is a first step to making sure that AI tools developed with federal funding at least have one check in place to mitigate harms those tools could cause to those populations. Existing practices in both harvesting data for AI and implemented that AI tend to reinforce existing power structures and replicate existing errors. Adding more people sensitive to these concerns into the pipeline is likely necessary but insufficient to address critics durable complaints about AI development and use.

Set to designate $2.2 billion to spend on AI over the next decade, the act stands to shape the AI field in a major way should it pass. Part of that will be through active effort, by establishing a National AI Coordination office and related committees to develop a strategic plan for AI research. Some of that will be through the National Institute of Standards and Technologies, by requiring it to create standards for evaluating AI algorithms and assess the quality of training data sets. The National Science Foundation will create educational goals and fund research through educational institutions, shaping the pipeline of AI workers for years to come. And the act will fund an AI research program in the Department of Energy, making up to five new AI research centers available primarily to government and academic researchers but open to private sector use.

Altogether, the act is about building the physical, human and financial infrastructure to sustain guide, and coordinate AI research, with the assumption that shared standards and the country housing the work will spill over into ethical use.

It’s worth noting that the AI-IA is distinct from Heinrich and Portman’s “Armed Forces Digital Advantage Act,” which seeks to improve the AI workforce within the Pentagon. AI is a field-agnostic technology, so efforts to fund research in the civilian world will likely have a spillover effect to military users, especially if the military has the skilled personnel on hand to take advantage of the new technology.

Like everything involving JEDI, there’s an extensive backstory. The cloud is essentially infrastructure, storage space that people and businesses pay for that allows them to access the same files from the internet, wherever they may be. That makes it somewhat different than traditional IT contracts, since it’s not exactly software and it’s not exactly hardware and, once adopted, it becomes the kind of thing future software and hardware depends on. That makes the contract itself, designed for a single supplier and an indefinite duration, worth $10 billion from the start.

Both the monetary value of the contract — and the intangible, inextricable way JEDI will be tied to future military operations — make it an important process to get right. The contract is expected to be awarded in April 2019, but it’s already drawn protest and legal action for how it is structured and for possible connections between the people awarded the contract and its likely recipients. Going single-supplier likely means granting the contract to either Google or Amazon, whose Amazon Web Services cloud houses much of the existing internet.

The Open the Government’s report, published March 13, examines Amazon’s existing contracts with parts of the federal government, including the Department of Homeland Security and the FBI. The report finds that, while privacy, transparency and accountability requirements for private contractors are already weak, they are especially so when it comes to Amazon and projects like JEDI. The nature of the cloud itself, and the logic powering algorithms and AI that run on cloud data, are inherently opaque.

To better understand the report, C4ISRNET spoke with Emily Manna, policy analyst with Open the Government and a co-author of the report. The below exchange has been lightly edited for clarity.

C4ISRNET: What prompted the report on Amazon?

EMILY MANNA: We started the research for this report by looking more broadly at private government contractors in the national security space, and it felt like all roads led to Amazon. Not only do they have cloud contracts with most federal agencies, they’re also pitching AI products to law enforcement and national security agencies that are generating lots of controversy. While the federal government is rife with revolving door and corporate capture issues, Amazon has garnered particular focus because of the JEDI contract process and HQ2. Amazon is also less transparency than many of its rivals, and works very hard to stymie public records requests.

C4ISRNET: Is there anything in particular about the JEDI contract that stands out as particularly unusual or ominous?

MANNA: The most noteworthy thing about the JEDI contract is its size, and the fact that it will go to only one provider. It certainly seems that former Defense Secretary [Jim] Mattis was enamored by what he learned from Amazon and Google when he visited their respective headquarters in 2017, and that Amazon’s advantage is bolstered by its ongoing cloud computing services for the intelligence community. The Oracle lawsuit has raised some serious ethical questions about potential conflicts of interest in DoD, and the Inspector General and FBI are now investigating whether the competition might have been unfair from the beginning.

C4ISRNET: How deep is the problem of a lack of transparency here?

MANNA: The secrecy goes very deep. In addition to the fact that complex AI systems are currently virtually immune to oversight because of the lack of explainability, the fact that much of this work is being carried out by private contractors like Amazon, Google, and other smaller firms harms transparency. Private companies don’t have nearly the same transparency requirements as government agencies, so a lot of the information about these contracts is protected as “trade secrets.” Finally, there’s just the larger problem of over-classification in the national security agencies.

C4ISRNET: Do we know enough to know what we don’t know?

MANNA: The short answer is “no.” We are almost completely in the dark about how the CIA and the rest of the intelligence community are using Amazon’s cloud computing services. We know only that CIA officials have bragged about the AI capabilities they have because of the contract with Amazon, but the possibilities for how they’re using AI are almost limitless, and some of those possibilities could be pretty scary if they have anything to do with counterterrorism and the drone program. DoD tells us quite a bit more about how they’re using and want to use this tech in the future, but we still don’t know much.

For example, we know generally about Project Maven and that DoD considers it to have been a huge success, but we don’t actually know anything about whether it actually worked. We know there were lots of civilian casualties resulting from U.S. airstrikes around the time Project Maven was operationalized, but we have no idea how AI may have played into those deaths.

C4ISRNET: This opacity seems fundamental to AI and Amazon. Are there some hard limits in how indecipherable AI can be? Would posting the code openly be a way to get oversight onto the functions of the AI?

MANNA: It certainly seems like open-source coding could help oversight bodies and the public better understand how government is using this technology — one wrinkle, of course, is the lack of technological expertise in Congress, due in large part to the fact that Congress just doesn’t have the resources to recruit expert staff and pay them competitively. But it does also seem like some of this may be inherent to complex AI and to working with private companies like Amazon. Congress should absolutely mandate that private contractors be more transparent to the public. When a private company is fulfilling government functions, there is a need for oversight and accountability that can’t just be written off for the sake of protecting proprietary information. Companies must also ensure that they know how their technology will be used before they sell it to government agencies, and that the technology is not immune to oversight. The federal government is understandably intent on staying ahead of the curve in the AI arms race, but they also need to be accountable to the public, and that means being able to assure that the technology they are using is predictable and explainable.

C4ISRNET: What of the likely counter-argument that transparency in, say, weapons AI is a security risk? Is this covered by the recommendation for a classified annex in reports?

MANNA: It’s, of course, true that some information will need to remain classified so as not to jeopardize national security. But right now, over-classification is a much bigger problem in the federal government, with experts estimating that between 50 and 90 percent of classified government information could safely be released to the public. Moreover, it’s possible that excessive secrecy about AI systems could be harmful to national security, because complex systems are often vulnerable to adversarial image attacks and other types of hacking. Many contractors providing services to DoD have already been hacked, including one of the AI firms that worked on Project Maven. Without robust oversight of these systems, there can’t be real accountability for these failures.

C4ISRNET: What should contractors know if they want to both win government, and especially Pentagon, AI contracts?

MANNA: Companies need to take responsibility for how their services will be used by the Pentagon and other national security agencies. Agencies should be able to give contractors information about the types of operations and programs their services will be used for, without giving away specific operational details. Contractors should not only release more information to the public, they should also stop trying to prevent the government from releasing information about their work to the public through FOIA, an unfortunately common practice.

C4ISRNET: What’s your sense of urgency in the timing of the report?

MANNA: The most urgent thing is that the time to begin regulating and demanding transparency about the military’s use of AI systems is now. The more AI becomes integrated into the day-to-day operations of the Pentagon, the harder it will be to determine where this technology is at work — and when it’s responsible for failures or abuses.

Wray, whose remarks came in conversation with Lawfare Executive Editor Susan Hennessey, opened by listing a variety of threats, first by the specific names of the bundles of code, and then broadly framing a picture of a dangerous and anarchic domain.

“We’re seeing an uptick in the threats from various foreign adversaries, and the blended threat, where foreign power enlists criminal hackers, essentially as mercenaries,” said Wray.

How can a state, specifically the U.S. government, adapt to and counter such a field of threats? By working closely with the private sector, Wray argued, noting that 90 percent of U.S. critical infrastructure is privately held and that any defense requires partnerships between law enforcement and business. Wray specifically mentioned that one way companies can do this by working with local FBI offices in the United States, and also with the FBI’s dozens of offices abroad, to proactively form relationships and work on mitigation, enabling the FBI to work closely with companies when a crisis comes.

Opt-in cooperation works when interests between the FBI and companies align. That’s not always the case and, in the past, the FBI has come under fire for seeking a way around cryptographic protections in cases where those interests don’t align.

“We’re not seeking backdoors any more than we think people on the other side are trying to actively weaken public safety,” said Wray, noting that there is much that cybersecurity protects on behalf of legitimate interests. But, continued Wray, we “cannot have a stable endstate where there is entirely unfettered space beyond the reach of law enforcement.”

Wray offered no specifics of how to get law enforcement into that unfettered space. Instead, the conversation turned to threats the FBI is actively monitoring and tackling. Wray noted that while the 2018 midterms in the United States saw no material impact on election infrastructure, they instead saw the continuation of a malign foreign influence campaign designed to sow discord and pit Americans against each other. How that differs from the default nature of political campaigns in the 2010s was also not addressed.

Wray emphasized a renewed focus on counterintelligence at the FBI, meant to counter specific espionage from China across every sector of the economy and institutions like academia, noting that counterintelligence efforts had been under-prioritized for too long. Wray noted that people were charged in conjunction with this counterintelligence effort, which prompted Hennessey to ask if the high-profile arrests had been driven by considerations of ongoing trade negotiations between the United States and China.

“The arrests were not driven by trade, not driven by politics, not driven even by diplomacy … they were driven by facts,” Wray said, to applause from the attendees. “If we find somebody committing federal crimes against American businesses, we’re going to go after them.”

That law enforcement is inextricably linked to rules set about trade, conditions produced by politics, and the roles performed by diplomats in a country were left undiscussed. Asked about the deterrent effect of indictments on hackers, Wray pointed to the impact being indicted had on specific bad actors.

“I will say that one thing that hackers of all shapes and sizes love most is anonymity; by indicting publicly you strip them of that,” said Wray. “Particular hackers have become personas non grata. They have been unable to find work.”

That’s a compelling case for deterrence on the individual level, though it doesn’t address the broader actions of states that institutionalize organs of cyber espionage, nor does it broadly change the dynamic of countries specifically contracting or enlisting criminal organizations to do their dirty work.

Wray may be plotting a path where the FBI can enter those unfettered spaces, but his answers on the floor of RSA spoke to a philosophy of individual capture and punitive expeditions. The problems facing a world built on cryptographic architecture are impossible to solve by law enforcement alone.

There are degrees and worlds in AI, a vast space from deterministic and emergent behavior, from online machine learning to targeting tools, and it is in that complexity that care is most required, that human direction is most desired, that the fears and possibilities of generations of science-fiction authors await to be realized. It is a complexity, in form and function, that is lacking from both the White House’s new AI strategy. For all intents and purposes, the document guiding the government’s approach to artificial intelligence might as well say “fancy software” and be done with it.

In the “Executive Order on Maintaining American Leadership in Artificial Intelligence,” the White House states that its efforts in AI are guided by five principles. Those principles are, roughly: driving technological breakthroughs; development of technical standards; training workers with new skills for an AI economy; balancing trust in AI and protection of civil liberties; and supporting “an international environment that supports American AI research and innovation and opens markets for American AI industries.”

Nowhere is the how or the what of AI spelled out. There is AI investment throughout the federal government, and the White House wants to make sure it continues in a permissive environment, but the term is an umbrella, a catch-all, with no specificity as to what it does, or why it might require efforts to train new workers.

“I applaud a number of aspects of the executive order, such as the proposal — mirroring the white paper I released last summer — to open federal data-sets to non-federal entities” said Sen. Mark R. Warner, D-Va., vice chairman of the Senate Select Committee on Intelligence. “Overall, however, the tone of this executive order reflects a laissez-faire approach to AI development that I worry will have the U.S. repeating the mistakes it has made in treating digital technologies as inherently positive forces, with insufficient consideration paid to their misapplication.”

Warner’s white paper lightly describes AI, highlighting the use of machine learning to train algorithms on pattern recognition in images, for example. But overall the paper is focused not on the risks of how AI is coded, but on the dangers that could come from a small segment of the technology industry capturing and dominating an emerging market.

If there’s an obvious precedent to all this, it’s rooted in the early 1990s. A confluence of factors, including a push from the Clinton administration, proactive legislation in Congress and favorable rulings from the courts created the conditions that would foster the Silicon Valley-led tech industry in its current form. Parallel to that initiative was getting the government itself to adopt the then-young technologies of the information age, all under the aegis of “reinventing government.”

While the White House is light on the specifics of what it wants AI to do, or even what exactly AI is, the Pentagon has a far more workable definition.

“AI refers to the ability of machines to perform tasks that normally require human intelligence — for example, recognizing patterns, learning from experience, drawing conclusions, making predictions, or taking action — whether digitally or as the smart software behind autonomous physical systems,” reads the Department of Defense AI strategy summary.

It helps that different parts of the Pentagon have been defining and working on AI in different ways. In June 2018, the Pentagon stood up the Joint Artificial Intelligence Center, or JAIC, and the military is already involved in the development and acquisition of specific AI projects. Project Maven, likely the most famous of these, was contracted through Google to adapt open-source tools to identify objects in drone videos.

“The impact of artificial intelligence will extend across the entire department, spanning from operations and training to recruiting and healthcare,” said Dana Deasy, the Defense Department’s chief information officer. “The speed and agility with which we will deliver AI capabilities to the war fighter has the potential to change the character of warfare. We must accelerate the adoption of AI-enabled capabilities to strengthen our military, improve effectiveness and efficiency, and enhance the security of our nation.”

The overall approach is focused on AI-enabled capabilities, using a shared and scalable foundation for AI, training an AI workforce, working with business, academia and allies, and leading the world in “military ethics and AI safety.” That last section is likely to attract standalone significance, especially since some companies have already made their future work with the DoD contingent to how it responds to questions of AI ethics

“We will invest in the research and development of AI systems that are resilient, robust, reliable, and secure; we will continue to fund research into techniques that produce more explainable AI; and we will pioneer approaches for AI test, evaluation, verification and validation,” the document reads. “We will also seek opportunities to use AI to reduce unintentional harm and collateral damage via increased situational awareness and enhanced decision support.”

The strategy is a promise to develop principles more than an outline of principles themselves. Of particular note is the notion of AI as a tool explicitly for the reduction of harm and collateral damage. That same language is often used in the explanations for the use of precision weapons, though the circumstances guiding the development of precision weapons were all about battlefield utility first. AI that can reduce collateral damage is also likely AI that can be part of autonomous targeting. To its credit, the DoD strategy specifically acknowledges the risk that might come from “’emergent effects’ that arise when two or more systems interact, as will often be the case when introducing AI to military contexts.”

Borrowing and adapting innovations from the commercial space for tasks such as predictive maintenance and supply could be part of a quiet logistics revolution within the military. Warehouses adequately stocked with exactly what is needed or parts moved to the bases and troops before shortages occur could sustain operations at reduced cost, as imagined by the strategy, or at a higher tempo, as is possible when commanders adjust to the new normal.

Indeed, one of the better case scenarios for the military and the government’s adoption of more AI technologies is simply adapting what has already worked in the private sector and with contractors to new missions. If there is funding and effort behind the new AI plans, it could realize many of these advantages first in cybersecurity.

“Today, AI is at the center of most major technological advances in areas as varied as cybersecurity, self-driving cars or development of cancer treatments,” says Dmitri Alperovitch, the co-founder and chief technical officer of CrowdStrike. “In cybersecurity, for example, these technological advances include enabling the defenders to recognize and stop never-before-seen attacks and being faster in detecting and responding to attacks.”

How, exactly, the United States fosters, adopts, and uses the tools powered and enabled by AI is likely of major importance for decades to come. The existence of national strategies for AI suggest that the government at least knows it needs to take AI seriously. Both the DoD and the White House seem content to follow the lead of the private sector, where there is a tremendous amount of work being done on everything from warehouse management to automated surveillance systems and innovative cyber security techniques. Yet if there is any overall ethos guiding the strategies, it is a need to meet the technology as it already is. Much of the government’s AI strategy is about playing catch-up.

Speaking before the subcommittee were Lisa Porter, deputy undersecretary of defense for research and engineering, and Dana Deasy, chief information officer for the Department of Defense. Central to the conversation over what, exactly, the Pentagon is doing with artificial intelligence were three big themes: making the data useful, making the data verifiable and making the data secure.

Money for something, algorithms for DoD

Key to turning AI from nebulous concept and flashy buzzword into a useful way to process information is funding, and while the overall budget for AI within the Pentagon is still unclear, there are some very real numbers for funding, of which Dr. Porter was happy to remind the subcommittee.

Over the course of the next five years, DARPA is set to spend a total of $2 billion on AI through a variety of initiatives. These include quick turnaround projects through the Artificial Intelligence Exploration program, which wants to get researchers started exploring a concept within 90 days of announcement and have some answer as to whether that AI concept is viable 18 months out from funding. This is broadly part of DARPA’s AI Next Campaign, which is also looking for long-term benefits and concepts for AI that are beyond what commercial companies are looking at for their own purposes.

Between the immediate term and the long-term funding sits the work of DIUx, which wants to see if there are tools and techniques already viable in the commercial world that could be adopted, off-the-shelf or with modification, into the operation of the Pentagon. JAIC, the Joint Artificial Intelligence Center, can serve to mediate those flows, looking for projects coming online in the near term and trimming redundancy where possible.

Data, not lore

What is especially compelling about machine learning is the ways it reaches useful conclusions while diverging from simple and strict deterministic algorithms. This makes an AI algorithm something of a black box, where it is fed data and then produces a processes result, which is hopefully something valuable downstream.

“Another well-known limitation of current systems is that they cannot explain what they do, making them hard to trust,” said Porter. “In order to address AI’s trust issue, DARPA’s Explainable AI program aims to create machine learning techniques that produce more explainable models while maintaining a high degree of performance.”

The opacity of AI involves two interrelated risks. The first is that people who need to use the data won’t trust it when it’s working correctly, and the second is that the opacity allows an adversary to mingle with the processing on the inside and produce a potentially dangerous result. This is especially true in the missions where AI interpretation can be life or death, which fits both the Pentagon and some commercial applications.

“People can spoof AI systems pretty easily,” said Porter. “One notable example is in the self-driving car community, [where] a team in Berkley put tape on stop signs and AI thought a stop sign was a 45 mph speed limit sign. There are countless examples of this now.”

Most of the actual fighting will be done by small robots

Beyond failures in funding and errors in processing, the other major threat to the United States’ ability to reliably use AI is losing an edge through the theft of data. To protect AI research from another nation-state’s thieves and saboteurs, Deasy says the JAIC is turning to a novel solution: “We’re going to apply AI to help us solve this problem.”

Pointing to some ongoing national and component mission initiatives, the JAIC is working with cyber command to see how to apply AI and pattern recognition to detect intrusions and anomalies in sensitive networks, and quickly assess if there’s been a change from normal behavior.

“If you think of how hackers normally try to penetrate, they’ll go to a point of least resistance, and once they’re in they’ll go laterally, and then what you’re looking for is exfiltration,” said Deasy. “We believe AI will be a very good machine use case for how we look at data patterns and signatures across our network to help ensure we don’t have exfiltration occurring to folks like the Chinese.”