Loading...
Defense policy is legislating who is accountable for AI-informed decisions. The IBM data shows CIOs already feel the gap. The CMO is in the same position and less likely to know it.
On June 8, 2026, Senator Adam Schiff introduced the Human Authority in Lethal Operations Act — a bill that would require the Department of Defense to designate a specific human commander with final say over any AI-enabled use of lethal force, with full disclosure of who that commander is and how the target was selected. The same week, IBM published a study of 2,000 technology executives finding that two-thirds of CIOs and CTOs are personally accountable for AI systems they do not fully control. Two unrelated stories. The same diagnosis.
The U.S. defense establishment has decided that the question of "who is accountable for the AI decision" is too important to leave to default. So has the most senior IT leadership in the Fortune 500. The marketing industry — and specifically the CMO function that nominally owns the strategic decisions an AI-augmented marketing organization is supposed to make — has not. The asymmetry is the most important structural problem in the buyer's market right now, and it is the problem an AI-native strategy agency is built to solve.
The HALO Act is unambiguous about what is at stake. According to UPI's coverage of the bill, the legislation requires the disclosure of designated commanders who authorize force involving autonomous weapons, mandates comprehensive records of the decision-making process including target selection, and requires a review process before any autonomous or semi-autonomous system is deployed. Schiff's framing is explicit: "we cannot depend on technology alone to guide us, particularly when the risks of harm can be fatal. My legislation would ... ensure that humans in the chain of command exercise responsibility for the use of any lethal technology."
The bill is not anti-AI. Schiff acknowledged that there are good reasons to use AI to advance national security. The bill is pro-accountability. The question it answers is structural: when an AI system contributes to a consequential decision, who is named, on the record, as the human who owns the call? Defense policy is now legislating the answer. The marketing industry is not even asking the question.
The IBM data makes the cost of the unanswered question measurable. The IBM Institute for Business Value surveyed 2,000 senior technology executives across 33 geographies and 19 industries between January and April 2026. Two-thirds of CIOs and CTOs report being held accountable for AI systems they do not fully control. Seventy percent of executives say teams across the business are deploying technology faster than IT can track. Seventy-seven percent say AI adoption is outpacing current governance capabilities. By 2027, surveyed organizations expect to deploy an average of 1,661 AI agents — a 38% increase — and only 11% of executives feel fully prepared for the scale of deployment coming in the next year.
The pattern is the same pattern the HALO Act is legislating against. The accountability sits at the top. The control sits at the edges. The gap between the two is widening every quarter as the agents ship and the governance does not.
The CIO is accountable for AI he or she does not fully control. The CMO is in the same position and is less likely to know it. The marketing organization in 2026 is operating with AI agents in nearly every function — content production, media buying, audience segmentation, competitive monitoring, brand listening, and customer journey orchestration. The CMO owns the strategic recommendation that those agents are supposed to inform. The CMO rarely owns the AI systems that produce the inputs to that recommendation. The CMO is accountable for a decision the CMO's organization cannot fully trace.
The structural parallel to defense policy is direct. The HALO Act does not ask whether lethal AI systems work. It asks who, by name, is on the record as the human who authorized the strike. The CMO's organization is making market-shaping decisions every quarter — where to compete, which segments to prioritize, which messaging to retire, which bets to double — and the human who owns the call is increasingly difficult to identify. The strategy is informed by a sprawl of AI tools. The recommendation is built by synthesizing fragments of insight from tools the team does not fully understand. The decision is made in a meeting where the data, the methodology, and the rationale are all presented as inputs rather than as accountable recommendations.
Tool sprawl is the symptom. The synthesis debt is the disease. The missing layer is the named human — or named partner — who owns the call. Defense policy is starting to legislate this layer. The CMO function has not yet recognized it is missing.
The IBM study points out that organizations deploying more AI agents faster are also experiencing more incidents. The average organization in the IBM sample reported 54 AI agent incidents in the past year that required human intervention or correction. Seventeen percent of those incidents were classified as high severity — more than four hours to contain. The breakdown of high-severity incidents is instructive: 37% resulted in data exposure or security breaches, 33% caused cascading system failures, and 17% triggered compliance issues. Organizations that embed governance directly into their AI systems experience 25% fewer incidents than those relying on manual oversight.
The marketing equivalent of those incidents looks different but follows the same pattern. A campaign ships with positioning that does not match the brand's strategic narrative. A media buy optimizes for a metric that the strategic team did not approve. A competitive brief gets circulated with a recommendation the senior strategist never reviewed. A product launch misses the segment the strategy was supposed to prioritize. None of these are the kind of incident a CIO would file in an incident-management system. All of them are the kind of incident that produces a quarter the CMO has to explain to the board.
The marketing organization is running the same AI deployment pattern as the rest of the enterprise — faster than governance, with more agents than oversight — without the formal accountability structure that defense policy is starting to require. The CMO is the CIO of the marketing function, and the marketing function is making decisions with the same control gap. The difference is that the CIO has a board asking questions about AI governance. The CMO has a board asking questions about the P&L.
The HALO Act answers three structural questions. Who is the human accountable for the decision? What was the decision-making process, and can it be reviewed? Was the system authorized for use in this context, and was the authorization reviewed? Translated into the marketing context, the questions are equally direct.
Who is the human accountable for the strategic recommendation? In most marketing organizations in 2026, the answer is the CMO — but the CMO's recommendation is built from outputs the CMO does not author and from tools the CMO does not operate. The accountability sits with one human; the work sits with many systems. The defense version of this problem is being solved by naming the commander on the record. The marketing version is being solved by no one.
What was the decision-making process, and can it be reviewed? In defense, the HALO Act requires comprehensive records of how targets were selected. In marketing, the equivalent is the strategic memo — the document that records what the recommendation was, what data it was based on, what alternatives were considered, and what assumptions the recommendation depends on. Most marketing organizations in 2026 cannot produce that memo on demand. The work is fragmented across dashboards, briefs, slack threads, slide decks, and the institutional memory of the team that built it.
Was the AI system authorized for use in this context, and was the authorization reviewed? The HALO Act mandates a review process for any autonomous or semi-autonomous system before deployment. The marketing equivalent is the question of whether the AI tools informing the strategy have been validated for the specific decision the team is making. Most marketing organizations treat their AI tools as if they were authorized by default. The assumption is that because the tool is in the stack, the tool is appropriate for the decision. The defense policy treats that assumption as unacceptable. So should the CMO.
The reason most marketing organizations cannot answer the three structural questions above is not lack of investment. It is the synthesis tax. The average marketing team is running somewhere between six and twelve separate AI tools — each one producing fragments of insight that need to be integrated, validated, and translated into a single defensible recommendation. The synthesis work falls on the team's highest-leverage people, who spend their time reconciling dashboard outputs instead of making decisions. The work product gets produced. The strategic accountability does not.
This is the same pattern the IBM study identifies at the IT layer. Organizations that embed governance directly into their AI systems experience 25% fewer incidents and deploy 16 times more agents than those relying on manual oversight. The marketing equivalent is the difference between a team that operates twelve tools and synthesizes the outputs manually, and a partner that absorbs the tool sprawl, owns the synthesis, and puts a name on the recommendation. The first produces fragments. The second produces decisions.
The defense policy answer to the control gap is structural: name the human, record the process, review the authorization. The marketing answer is relational: name the partner, document the synthesis, own the recommendation. The structural logic is the same. The marketing version does not require legislation to enforce — it requires the buyer to demand it.
Cannes Lions opens on June 22. The festival will be the year's largest concentration of "AI consultancy," "agentic AI," "AI workforce," and "AI platform" announcements. The 4As/DBC report on entry-level agency positions landed the same week as the IBM study and the HALO Act — and made the same point. According to MediaPost's coverage of the report, 4As CEO Justin Thomas-Copeland warned that AI has "hyper-accelerated" the compression of the entry-level agency pipeline "much more aggressive than in any sort of disruptive time that" the industry has previously navigated. The pipeline that historically trained the next generation of strategists is being hollowed out by the same automation the platforms are selling. The synthesis tax is being pushed onto senior people who do not have the bandwidth to absorb it. The strategic accountability layer is thinning at exactly the moment the buyer's environment requires it to thicken.
Three diagnostic prompts separate the platform pitch from the strategic partner, and they map directly to the questions the HALO Act is trying to answer for defense.
First, who is the named human accountable for the strategic recommendation this engagement produces? If the answer is "our platform produces the analysis and your team decides," the buyer is buying infrastructure. If the answer is "we own the recommendation and we will defend it on the record," the buyer has found a partner. The defense version of this question is now law. The marketing version is still optional. The buyer should not accept optional.
Second, can the partner produce the strategic memo on demand — the document that records what the recommendation was, what data it was based on, what alternatives were considered, and what assumptions it depends on? If the answer is no, the partner is selling access, not accountability. The HALO Act equivalent is the comprehensive record of how targets were selected. Most AI services cannot produce that record. The strategic accountability layer requires it.
Third, was the AI system validated for this specific decision, and was the validation reviewed? The HALO Act requires a review process for any autonomous or semi-autonomous system before deployment. The marketing equivalent is whether the tools producing the strategic recommendation were chosen for the question being asked, rather than inherited from the existing stack. Most marketing engagements treat the tool as the answer. The strategic accountability layer treats the question as the answer.
The Defense Department is legislating human authority over AI decisions because the stakes of unaccounted AI are too high to leave to default. The Fortune 500 is discovering, through the IBM data, that the IT organization is already accountable for AI it cannot fully control. The marketing organization is in the same position and is less likely to know it. The buyer is making decisions with the same control gap the defense establishment is trying to close. The Cannes announcements in two weeks will not address this. The platform pitches will not address this. The AI consultancy branding will not address this. The only layer that addresses this is the strategic accountability layer — the engagement model in which a partner owns the recommendation, defends it on the record, recalibrates when the market moves, and is named as accountable when the call turns out wrong.
That is the layer an AI-native strategy agency occupies. The same tool sprawl the rest of the industry is consolidating and automating, an AI strategy agency absorbs. The same synthesis tax the marketing team cannot carry, an AI strategy agency carries. The same accountability question the HALO Act is legislating for defense, an AI strategy agency answers for the marketing function. The tool produces fragments. The partner produces decisions. One subscription. The accountability is not an add-on. It is the product. End the sprawl. Get outcomes.
Ready to move from platform pitches to strategic accountability? Get started with Autostrat.
Book a 30-minute demo. Bring a live question and watch the answer get built.