The Federal AI landscape is enormous, and the terrain varies widely. Whether you’re entering for the first time or hiking in a new area, DWPA’s AI Innovation Cell and AI Landscape report are the map and compass you need to avoid missteps and wasted time.

There’s no standard way organizations first try generative AI, but it’s common for early adopters to use it, initially, on job-related tasks. One user searches for specific experience in scattered resumes. Another analyzes and formats data for a report. A third revises content for a proposal. Early adopters typically set out “to see what they can do” using tasks they know well, and then see what they learn.

On October 30, 2023, the Biden Administration issued its Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. You can read DWPA’s summary of the Order’s purpose and intent here. Below we explain what the Order’s language about “dual use technologies” could mean to your business.

Artificial Intelligence has been part of government contracting and consulting for decades. As of September 1, 2023, AI.gov lists more than 700 use cases across 19 departments. The U.S. Government Accountability Office’s December 12, 2023 report, Artificial Intelligence: Agencies Have Begun Implementation But Need to Complete Key Requirements, identifies more than 1200 current and planned uses in 23 departments. And the General Services Administration identifies over 1200 members from 60 agencies in its AI Community of Practice.

Generative AI (GenAI) promises to increase use cases as readily available tools make GenAI accessible, affordable, and powerful for government agencies and contractors.

The Biden Administration’s Executive Order renewed focus on how AI policy will impact competitiveness, intellectual property, privacy, and national security. A key impact for U.S. companies will be compliance with Export Controls as firms consider export constraints to develop, implement, and offer AI systems and tools. 

Key Things to Know

Robust export controls already exist in the US, in two ways. One is “defense articles and services” governed by the State Department’s International Traffic in Arms Regulations (ITAR). The other is control of “dual use” technologies with both commercial and potential national security uses, governed by the Commerce Department’s Export Administration Regulations (EAR).

It’s relatively straightforward to identify and apply controls to “defense articles and services” subject to ITAR. It’s in the area of dual use technologies that regulations are less well-known, and more ambiguity exists. These require vigilance on the part of companies to ensure compliance as they consider how to employ AI in their offerings.

A critical business question is what will be controlled?  Generally, dual use technologies are controlled by “item-based” controls like systems and hardware (e.g. CHIPS Act export direction impacting advanced semi-conductor release), or by “end-user” controls on countries, organizations, or individuals (e.g. the “Entity List”). But there is also a category of less well-understood controls that focus on the “end use” itself, and place obligations on exporters to have “knowledge” of what end users might do with the technology. These are end uses that could involve support of nuclear, missile or unmanned aerial vehicles, or chemical/biological capabilities.

The responsibility to abide by these controls and requirements for compliance is entirely on “US persons,” defined as both individuals and companies. There are substantial penalties, both criminal and civil, that apply to both.     

Call To Action For Companies

In addition to the existing export controls, the EO will almost certainly drive new rulemaking at both the Departments of State and Commerce. Future regulations combined with the fast-evolving AI landscape mean companies should carefully evaluate and address export controls as they bring new capabilities to market.

The best practice is normally to obtain expert export control and/or legal advice. The best risk management for companies at this point is to do this early to avoid risky and potential expensive impacts from export considerations.

To learn more contact Lou.Kerestesy@DWPAssociates.com.

On October 30, 2023, the Biden Administration issued its “Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence.”

Deep Water Point & Associates (DWPA) has substantial experience with laws, regulations, guidance, programs, and requirements central to the Order. We’re analyzing the EO to understand whether and how it might affect our AI and generative AI use, and our clients. Federal contractors and SaaS or PaaS cloud service providers not well-versed in Department of Commerce Export Administration Regulations (EAR) might start investigating. This article summarizes the Order’s purpose and content.

This EO builds on recently published AI strategic documents and frameworks by Federal agencies and institutes. It points out that much of what’s already done for software development and data laws also applies to AI. With 186 shall statements and 98 deadlines, the EO establishes clear direction and cadence for next steps needed by the Federal government. It addresses AI and generative AI, and clearly describes when export control law and regulation apply.

The Order’s 13 sections are summarized below. Sections 4 – 11 constitute the Order’s “eight guiding principles and priorities.”

Sec. 1. Purpose emphasizes the significance of responsible AI use, highlighting its potential to address critical challenges and improve various aspects of society, while also acknowledging the risks associated with irresponsible use. It underscores the need for collaboration between government, the private sector, academia, and society to harness AI for good while mitigating its risks.

Sec. 2. Policy and Principles states that it is the policy of the Biden Administration to advance and govern the development and use of AI in accordance with eight guiding principles and priorities. This Section separately describes the eight guiding principles and priorities, which are Sections 4 – 11.

Sec. 3. Definitions defines 32 terms. Noteworthy among them is the term dual-use foundation model, which is used 16 times and is central to developer, user, and agency requirements and prohibitions.

Sec. 4. Ensuring the Safety and Security of AI Technology is the largest of the Order containing more than a quarter of the entire Order, one-quarter of its deadlines, and almost one-third of its shall statements. This section details guidance and direction pertaining to safe and reliable use, almost two dozen infrastructure as a service requirements, cybersecurity, biosecurity, and other types of uses and risks. Section 4 contains one of two uses of the term red-teaming pertaining to generative AI. Section 10 contains the other.

Sec. 5. Promoting Innovation and Competition outlines measures to attract and retain AI talent to the US, promote innovation through public-private partnerships, provides guidance to patent examiners, and identifies measures to support AI in healthcare, for Veterans, in climate change, scientific research, and other domains.

Sec. 6. Supporting Workers emphasizes the government’s commitment to understanding and addressing AI impacts on the workforce. It directs the development of reports analyzing labor market effects and workforce disruption mitigation principles and best practices, and education and workforce development.

Sec. 7. Advancing Equity and Civil Rights outlines the government’s efforts to address discrimination, promote equity, and protect civil rights in various aspects of AI deployment, including the criminal justice system, government benefits and programs, and the broader economy.

Sec. 8. Protecting Consumers, Patients, Passengers, and Students highlights the government’s efforts to ensure the responsible and ethical use of AI in healthcare, education, transportation, and communications, while protecting consumers and addressing potential fraud, discrimination, and privacy risks.

Sec. 9. Protecting Privacy emphasizes the government’s efforts to address and mitigate privacy risks associated with AI, promote the use of privacy-enhancing technologies (PET), and support PET guidelines, research, and development.

Sec. 10. Advancing Federal Government Use of AI is the second largest section of the Order. It highlights steps and guidelines to advance the Federal government’s use of AI and enhance its AI talent and management. It forms an interagency council to coordinate the development and use of AI in agency programs and operations, other than the use of AI in national security systems. Section 10 contains all the Order’s only references to the Technology Modernization Fund. It also contains one of two uses of the term red-teaming pertaining to generative AI. Section 4 contains the other.

Sec. 11. Strengthening American Leadership Abroad underscores the importance of the United States in global AI leadership, setting standards, promoting responsible AI development and deployment abroad, and addressing cross-border AI risks, particularly in critical infrastructure.

Sec. 12. Implementation establishes the White House AI Council which will coordinate AI-related activities and policies across the Federal government. It identifies the Assistant to the President and Deputy Chief of Staff for Policy to serve as the Councils’ Chair. It identifies 28 agencies’ secretaries, directors, and chairs as members, plus the heads of such other agencies, independent regulatory agencies, and executive offices as the Chair may designate or invite to participate.

Section 13. General Provisions ensures that this EO is not read as impairing authorities granted by law, or as establishing existing authorities or government functions.  

To learn more contact Lou.Kerestesy@DWPAssociates.com.

A conference presenter recently told an audience, “Whatever you put on ChatGPT is out there. Gone for good. Out of your control.”

We hear that dire warning a lot and it raises serious concerns about business use of public tools like ChatGPT or Bard. The warning could also be more cautious than it needs to be, and cost you more than it buys in protection. Let’s see.

What Is Generative AI, And How Does It Work?

Most software we use is deterministic. It produces the same output given the same inputs and conditions. We rely on that predictability when it comes to writing emails and reports, and analyzing sales or budget scenarios.

By contrast, GenAI is generative. It’s designed to produce diverse and even creative outcomes using the same or similar inputs. We want it to brainstorm with us. To summarize a report in its words. Or to change the tone of an email for us.

GenAI does this by using language patterns. It recognizes the relationship of words, phrases, and sentences and then uses statistical probability to select the best sequence of words to return to you, based on your prompts.

When you hear talk of GenAI training, this is what’s meant – training it to recognize and use language patterns. As an example, ChatGPT was trained on 300B words, including scoring and weighting them based on how they were used in sentences. This “deep learning” is what makes generative AI useful.

What Has GenAI Training to Do with Safe Use?

The way GenAI works tends to limit what others can know about your use. While it’s true that GenAI tools read your prompts and might store them for future training, GenAI’s focus on language patterns rather than whole entries helps control risk but not eliminate it. Consider an example.

Say you cook and want to make a tomato sauce you’ve never made before. You search online for something you haven’t heard of, and search engines return entire recipes to you. All the ingredients, quantities, steps, and times for you to read – as you would expect.

But what if you used GenAI?

Let’s say I had previously put my grandmother’s secret tomato sauce recipe – which includes a dash of soy sauce at the end – in a prompt asking a generative AI tool (a GPT) to make a shopping list for me. Let’s also say the GPT stored my prompt for future training. Would it return my grandmother’s recipe to you like search engines would?

Because GPTs analyze language patterns to return language patterns to you, it’s not likely to return her entire recipe the way a search engine would. But, had you told it you wanted to try something unusual, it could very well inform you that “Some tomato sauce recipes use a dash of soy sauce at the end” because that’s novel. It could offer that tip along with others, all based on novel ingredients from thousands (tens of thousands?) of tomato sauce recipes.

It matters little if a GPT returns my grandmother’s entire recipe to you if her secret ingredient is identified for you. Her secret is out. But had you asked a GPT for Indian tomato sauce recipes, or different recipes with paprika, it might not have considered a dash of soy sauce at the end relevant. Remember, it’s all about what you ask and the relevance a GPT determines using language patterns and statistical probability.

So, is your proprietary or privileged business information at risk of being made public, through your use of GPTs trained on your prompts?

The answer is not no, but is it ever? The answer is yes, depending, and now you understand why. What, then, are safe uses of public GPTs?

A Word About Types of GPTs

AI terminology can be confusing. Glossaries contain dozens of terms, many of which sound like they say the same thing. Even the boundaries between simple terms like open, public, and proprietary aren’t so clean that certain terms always and only apply to ChatGPT or Bard, for example, while other terms always and only apply to, say, ACME Inc’s AI-assisted proposal tool. For the sake of easy reference, let’s divide products this way:

• Public refers to ChatGPT, Bard, and others you can try for free by registering at the tool’s website
• Private refers to dedicated, domain-specific tools you pay to use by user, per month, or by some other unit

We realize this might confuse architectures, fail to account for products with free and paid versions, ignore distinctions between publicly and privately held companies, and more. That’s okay because making those distinctions won’t change what we’re saying about safe use.

One safe-use advantage of a private tool is you can build and separate your document repository, and use only your repository to train the tool. Your vendor’s tool might also have a data relationship to foundational models, however, which might expose your data to others through training. Vendors know how to firewall your data and let you opt out of model training. Read the vendor’s and data use and privacy policies, understand the tool’s settings, and talk to the vendor if you have questions.

Can you also use a public tool safely? You can.

First, public tools might also permit you to prevent sessions from being used to train the GPT. Read their data use and privacy policies to understand how your data will be used, and to see if you can opt out of training.

Second, many valuable uses will have nothing to do with proprietary or privileged data. A proposal manager might use a GPT to improve their understanding of technical issues, to improve their conversations with technical SMEs. A team lead might role play with a GPT to understand the perspective of others on the team without ever using proprietary information. If you want to keep the risk-reward scales tipped in your favor, clarify what you want to accomplish with a particular use, know what success looks like, and ask yourself what might go wrong. You’ll find many ways to prompt a GPT which don’t require business data or information.

So, What’s the Bottom Line?

Recall my colleague’s dire warning at the conference: “Whatever you put on ChatGPT is out there. Gone for good. Out of your control.”

It’s true that the content of your prompts can be out there, depending on policies and settings. But it’s also true you can prevent the leaking of proprietary and privileged information.

But it’s also true that the way GenAI uses what’s out there reduces some risk for you. How safe that feels is a subjective judgment we’ll talk about in the next article. But understanding how GenAI trains helps you understand how information you provide in prompts can show up for future users.

In the GenAI Discovery Project, DWPA is experimenting with public and private tools. Using public tools, we know there’s zero chance we’ll give competition any advantage – because there’s no advantage at stake. There’s no soy sauce in the prompts. For uses where there’s a chance we could give something away, we know it’s a small chance and we weigh the gain we want from the harm we don’t want, and act accordingly.

DWPA has not used private tools, yet, beyond Discovery Project trials, so we can’t speak to practices with them. We know private tools have additional safeguards built in. If you use or are considering a private tool, talk to your vendor about how it’s trained and how your data might be included.

Whether using a public or private tool, read your tool’s privacy policy or statement. They’re not generally written for human reading, but gut it out so you know what’s happening to your data. You’ll probably see a choice for opting your content out of tool training. DWPA has exercised that option.

Beyond understanding how GenAI tools train and work, safe use comes down to use cases and risk tolerance. We’ll look at that in the next article but, for now, we’ll leave you with the thought that you probably already engage in a practice which is like determining GenAI safe use: Asking questions at an industry day, or in written Q&A during a solicitation process.

You can ask in ways which show your hand, or in ways which don’t. You weigh the odds of gaining information to your advantage versus benefiting your competition and neutralizing your gain. You might have done this for years, and it’s a risk-reward decision similar to deciding how to use GenAI, especially public tools.

To learn more, contact Lou.Kerestesy@DWPAssociates.com.

Part 1 of this two-part article defined adoption and talked about what makes it hard in any organization. Part 2 describes the ways you can manage adoption challenges.

Exec Summary: Over the past year, countless blogs, articles, books, videos, courses – even job descriptions – focused on prompts and prompt engineering. While prompting is essential to effective GenAI use, it’s only one thing to consider. Generative AI outputs are another, and they need more attention.

Recall the IPO Model – inputs, processes, and outputs. For a simple generative AI use, prompts are the inputs, algorithms are the process, and a GPT’s response is the output. For more complex uses, inputs and processes combine as a user and the GPT interact through a set of prompts. Outputs can also take on a new importance, depending where they lead.

If the outputs of my GPT use become inputs to a business task or process you own, we face added requirements for communication, collaboration, and probably change management. And that calls for an approach to addressing the hard adoption questions.

How Do We Answer The Hard Adoption Question?

Meeting the challenge of generative AI adoption will require a comprehensive and methodical approach. Here are three principles we’re applying at DWPA we recommend you consider.

1. Use an adoption framework
2. Clarify goals and objectives
3. Think like an entrepreneur

Use An Adoption Framework

The grand-daddy of innovation adoption frameworks might be Everett Rogers’ Diffusion Innovation Theory. In his 1962 classic (updated through a 5^th edition in 2003), Rogers explains how an innovation diffuses through, or is adopted by, a social system. There’s a lot to Rogers’ research and it would be worth your time to read select portions of this book. But we can highlight the pieces you can use immediately.

Most well-known might be adoption types Rogers identified and arrayed temporally in an S-curve. Rogers called it the innovation adoption customer because he studied many types of innovation. Today it’s popularly known as the technology adoption curve, as shown here: