Adopting a unified response to national security threats

A dominant theme at this week’s DoDIIS 2004 conference is unity of response, and for good reason: threats to U.S. national security and global stability are coming from multiple domains—cyber, economic, military, technological, and more. Malign actors are careful to avoid triggering us into a full united response until a time and place of their choosing.

However, those threats and the careful strategy behind them leave evidence in data, and this evidence can be picked up with increasing effectiveness, especially as they are emboldened to take more risks. In this post, I want to amplify my #1 takeaway for a broader audience: Their success requires our complacency.

Threats need us to be small-minded. This was a big topic in the DoDIIS day one fireside chat with Dr. Stacey Dixon and Gen. Dimitri Henry, moderated by Doug Cossa. They referenced 2022 remarks by Secretary of Defense Lloyd Austin that we are in a decisive decade, and we need to wrap our minds around that continually. At the end of their panel, they named one overarching risk—and opportunity: the community needs to take a “big E” Enterprise posture towards threat intelligence. Specifically, this includes leveraging more Generative AI (GenAI) to help analysts in using open-source intelligence (OSINT).

I’d like to propose a strategic approach to that risk/opportunity to continue the dialog beyond DoDIIS, and I’ll be hitting on this theme regularly on my LinkedIn. Please join!

Combining ontologies, GenAI, and OSINT to go faster

A common theme in other panels was the need to accelerate acquisition. One packed session briefed Project Higson, a 6-year research effort into semantic ontologies to describe two threats in the fullest possible detail. Once analysts built, validated, and aligned each ontology to intelligence questions rooted in a common core ontology, they used OSINT and GenAI to connect the dots faster than ever. These successful research efforts led to the creation of a new funding line for further exploration from FY25-FY30. I expect this to succeed and become a template for others.

Local success paved the way for “big E” Enterprise change. By combining ontologies, GenAI, and OSINT, we can enhance intelligence processes and present a compelling case to the “big A” Acquisition community for faster action and fewer hurdles to procure more ontologies, OSINT, and GenAI.

This is actually how Federal agencies go fast. The DoD Acquisition system is designed for large muscle movements on consensus threats, not workarounds and one-off contracts, as Dr. Dixon noted on the main stage. But there must also be proven successes to build on. OSINT, not considered a true collection discipline 20 years ago, has proven to be a rich source of intelligence “signal.” While OSINT also has noise, commercial technology can sift through it to illuminate threats. OSINT results are readily shared, enabling the community to accelerate efforts and protect more exquisite forms of intelligence. This is how intelligence cultures find unity of purpose and evolve into a “big E” Enterprise.

Building better intelligence with data governance and provenance

As more decision makers think big about these threats, they can require combinations of ontologies, OSINT, and GenAI to keep their specific missions relevant in this age of AI. The shift is well underway. The new National Security Memorandum (NSM) on AI and the DIA OSINT Strategy 2024-28 provide key principles and pathways for Enterprise elements to come together and move faster in concert to test, validate, and acquire what they need, without being bogged down by the overly restrictive centralized controls that our system protects when we are not under serious threat. But now we are, giving us the opportunity to move fast and at scale in a new unity of response.

Efforts by DoD’s Chief Data and AI Office (CDAO) have created momentum for this, too. Deputy Secretary of Defense Kathleen Hicks has shaped CDAO to leverage more of Silicon Valley’s innovation alongside the traditional defense industrial base, for times like now. New entrants like Seekr are incentivized to join the DoD innovation fray by investing in GenAI development toward becoming Awardable on CDAO Tradewinds Marketplace. Bringing our capabilities forward would have been nearly impossible without alternative acquisition paths like this.

Now, big Acquisition can reap these benefits to position small scale wins for big Enterprise data governance and provenance benefits, which is exactly what the nation needs now. This is not obvious but clearly implied by the NSM and DIA OSINT Strategy. I’d like to tease that out and show how the right kind of GenAI can build better intelligence products for the Enterprise by aligning (but not confusing) data governance and provenance processes.

This innovative approach to data management builds on operational wins like Project Higson, and it also helps the business of DoD by reducing Acquisition risk. Dr. Dixon and Gen. Henry agreed that the slow pace of acquisition is a major reason why DoD intelligence often resembles a “small e” enterprise, appearing patchwork, parochial, and stove-piped to its critics. Now, intelligence and acquisition analysts can use validated GenAI agents to do lower-level information tasks to move faster, with greater accuracy, and assured conformity to policy requirements.

I’ve personally heard many acquisition professionals lament that they want to speed up, too, but without a unity of response at all levels of government, their rules are burdensome without more hands on deck to do the work. More trusted AI agents can help with that.

The key is the alignment to policy requirements, and this starts, I believe, with knowledge of how our democratic political system and Cold War-era acquisition enterprise work and why they will not change on their own. Again, it comes down to unity of response. The big acceleration mechanisms are not available except in wartime, a point not lost on adversaries. The only alternative is for our community to come together, with the right knowledge base guiding our actions, and more AI agents on the field to help do the work.

The SeekrFlow™ enterprise AI platform can make a difference in this in-between period of less-than-wartime digital mobilization. By pulling governance and provenance principles out of these guidance documents, we can help automate the validated intelligence processes that the enterprise needs to ramp up acquisition—without sacrificing privacy and security. We help put more agents on the field to be managed with full transparency.

Aligning GenAI models to responsible AI governance

The Memorandum calls on the U.S. Government to cultivate “a stable and responsible framework to advance international AI governance that fosters safe, secure, and trustworthy AI development and use; manages AI risks; realizes democratic values; respects human rights, civil rights, civil liberties, and privacy; and promotes worldwide benefits from AI.”

SeekrFlow’s Principle Alignment technology can tune GenAI models to these values and principles. If we think of this international AI governance framework as one cultivated ontology, it can deliver next-level transparent governance and provenance for leaders charged with these solemn duties. Additionally, it offers data pipelines for analysts and operators who are seeking validated GenAI models. These models can be further cultivated with OSINT and automation to enrich their intelligence work and deliver the best options and insights to leadership.

To do this in collaboration with a wide range of allies and partners and still be trusted by American citizens wary of “international AI governance,” Principle Alignment builds the right links in the chain of AI governance.

The Principle Alignment process aligns recommendations and other AI-generated outcomes to legally defined oversight responsibilities, so Federal components can act with confidence under a shared operating model toward common goals. It also establishes provenance at the token (word) level, enabling strict oversight of the data feeding AI models, how the model is generating inferences from the inputs, and whether it comes from trusted sources that can be validated. All of this can be scored to de-risk the data flowing into large language models (LLMs) before they get too far into the intelligence process.

Navigating the path forward: challenges and opportunities in AI governance

This approach places GenAI and each LLM within a far more transparent and accessible risk framework, but significant challenges remain. Who gets to cultivate the framework in our polarized political condition—just one party or leader? Who decides which international partners can influence how we govern AI? How should we address emerging technologies that push the enterprise to move even faster? What is the right pace to thwart and deter threats? And how do we train young talent for these roles and incentivize them for the mission as their peers join slicker firms?

Still, it’s progress—and much needed. If we examine the DIA OSINT Strategy 2024-28 closely, we find a strong starting point on page 11, specifically under the section titled ‘Synchronize Across All PAI/CAI.’ While each use case is different, governance and provenance can be aligned to enhance synchronization across a broader array of decisions. This can be supported by more responsive data pipelines, rich in validated OSINT and transparently governed using Principle Alignment technology.

This is a good starting point to cultivate sound governance and drive a unified response to the threats now becoming openly manifest in public data.

Three guiding principles for integrating GenAI and OSINT into a strategic governance framework

I recommend agencies adopt three guidelines to link tactical PAI/CAI OSINT explorations to a broader governance strategy. Coming straight from GenAI, these themes encourage communicators to think and speak in more AI-oriented ways, which is critical for cultivating an AI framework that enables us, our allies, and partners to counter threats together in a shared spirit of strength for peace, prosperity, and inclusion:

1. Outcome-oriented interoperability

Gen. Henry noted that interoperability between teams and systems is an imperative; otherwise, “we hamstring ourselves in the long term.”

GenAI is outcome driven. OSINT data pipelines that feed GenAI configured for Principle Alignment will inherently enhance the interoperability of data and technology infrastructure. This is because common governance and provenance principles are used by GenAI to optimize intelligence outcomes, rather than merely protecting data stovepipes.

2. Risk-aware divestments

GenAI and OSINT—including both PAI and CAI—expand the variety and depth of insights available to the community. The best signal is often in public data. Efforts like AFRICOM’s Project Higson show how fruitful OSINT can be in illuminating relationships and transactions of concern in emerging risk areas.

While this may create “new work” to govern OSINT data sources under a robust provenance construct that limits what GenAI can consume, it doesn’t have to resemble a strict compliance process like the Risk Management Framework (RMF). Instead, this work can be done using ontologies that allow teams to focus data provenance work on enterprise value, like what connects the most dots in the ontology.

This is a sensible, risk-aware approach to bring more OSINT data feeds into intelligence systems. There is also a net acquisition benefit: data sources and stove-piped systems that do not prove trustworthy or valuable can undergo divestment analysis, allowing funds to be reallocated to more trusted, valuable sources that drive efficiency and accuracy.

3. Transparency-driven teams

Gen. Henry and Dr. Dixon agreed that informed teams who know their own gaps are best positioned to overcome them. Moreover, leaders do the most important work in modernization by refusing to withhold information about the real problems surrounding their legacy systems and operations.

They also challenged the assumption that only top leaders need the full picture. Today’s missions require more team members to have access to the complete intelligence landscape in order to identify the true risks facing their organizations. This allows them to problem-solve, think holistically, and make more informed contributions than if they were not “read in.”

The tendency to withhold information about system gaps makes sense on the surface—why highlight problems? However, adopting GenAI with more transparent governance and provenance processes expands the need to read in more folks, because GenAI is naturally outcome oriented. Over time, such transparency will accelerate teamwork to accomplish shared goals.

Building a unified response through governance and provenance

Successful digital modernization demands and rewards collaboration. GenAI’s natural outcome-driven orientation allows us to bring more AI vendors, compliance experts, security specialists, and mission owners into discussions that were once limited to top policymakers and elite intelligence analysts.

SeekrFlow’s Principle Alignment adds the force multiplier of collaboration with trusted AI agents. Aligned to the NSM and DIA OSINT strategy to start, and learning from examples like Project Higson, we don’t need to wait for more evidence; decision makers can act now to align Enterprise and Acquisition efforts for the pace and scale of today’s threats, without giving ground on our principles. In fact, they are what will make all the difference.

For strategic reasons, I appreciate how intentional the DoDIIS organizers have been about pushing this dialog into the open. By doing so, we give our adversaries the opportunity to pack it in and join the work of principle alignment (minus the autocratic vanity projects). There are too many human challenges at stake to waste time and money on vain things like AI dominance. While some aspects of political history may tempt leaders to reach for total information control, the success of the American spirit of innovation has never subscribed to the fatalism that says things can’t change. And the national strategy is far better for that.

That doesn’t mean we don’t need to change. On the contrary, this is not the Cold War, and strict Cold War-era controls do not help us build the right data and AI infrastructures the nation needs today. Instead, more transparent agents making sense of OSINT data feeds, validated by GenAI for informed decision-making, offer a better way forward.

I believe this approach answers the call for a unity of response from intelligence and military leaders like Dr. Stacey Dixon and Gen. Dimitri Henry. They know we need a calculated shift from legacy frameworks to AI-enabled infrastructures to maintain a sustained decision advantage, and they are looking to industry to think bigger to help.

With SeekrFlow, an enterprise component can code governance and provenance principles into AI systems with Principle Alignment technology, informed by proven ontologies. This ensures that AI-driven analysis and exploitation processes—and the resulting DoD intelligence Enterprise and Acquisition systems—are transparent, secure, and aligned with mission-critical requirements. After all, isn’t that how we win?