OpenAI Spends Millions on PR While Fundamental Issues Remain
This week, OpenAI released a 13-page document titled Industrial Policy for the Intelligence Age, bought a tech-aligned podcast network called TBPN, and announced an office in Washington D.C. dedicated to helping lawmakers and nonprofit organizations 'learn about their technology.' All this while surveys continue to document a growing trend: public disapproval of artificial intelligence is on the rise.
At a surface level, these moves appear to be those of a mature company doing what mature companies do: hiring influence, cultivating a narrative, and managing perception. However, a strategic reading reveals a more uncomfortable reality.
When the Message Becomes the Product
There’s a significant difference between communicating value and creating it. OpenAI is doing the latter, and the market will eventually recognize this distinction.
The recently published industrial policy document advocates for a 'reimagining of the social contract' with ideas that they themselves describe as 'people-oriented.' It’s the kind of language that sounds good on a panel in Davos yet changes nothing for a user whose job has been displaced or for a creator whose content was used to train models without their consent. The gap between what the document proclaims and what surveys record is not a communication issue; it's a value architecture problem.
What is strategically revealing is not that OpenAI is investing in narrative, but what that investment indicates about its competitive position. Companies with a solid value proposition do not need 13 pages to explain why they deserve to exist. Those that do are usually responding to pressure that their products have failed to resolve. The acquisition of TBPN reinforces this reading: when you purchase access to an audience rather than earning it through your product, you’re implicitly acknowledging that your product does not generate sufficient organic engagement.
The bet on Washington D.C. has a different logic, but is equally symptomatic. A physical space for lawmakers to 'discuss the company’s technology' is, in practical terms, lobby infrastructure with better interior design. This isn’t bad per se—any industry with regulatory impact needs a presence in political decision-making centers—but positioning it as a gesture of openness and education when surveys already show active distrust is risky. Legislators also read surveys.
The Variable No One is Eliminating
The structural problem in the AI industry right now isn’t public perception; it’s that most companies in the sector are competing on the same variables—processing speed, parameter volume, multimodal capacity—while systematically ignoring the variables that matter to the segments that have yet to adopt these tools.
Disapproval surveys are not random noise. They are signals of unmet demand. Entire segments of users, professionals, and organizations would likely adopt AI tools if the proposals came with transparency regarding training data, compensation mechanisms for original content creators, legal guarantees about copyright, and a learning curve that doesn’t require a PhD in engineering to calibrate results effectively.
None of these variables appear in OpenAI’s industrial policy document. What does appear is a call to 'reimagine the social contract,' which is a fancy way of asking society to adjust its expectations instead of adjusting the product itself.
Dr. Rebecca Swift from Getty Images articulates this succinctly from the visual content trench: when everything starts to look the same, audiences stop paying attention. This isn’t just an aesthetic issue; it’s a retention problem, and retention is the engine of any subscription model or data platform. The homogenization of AI output is not a bug; it's the predictable result of optimizing for speed and scale without sacrificing anything on the cost side. The industry’s response so far has been to produce more output, faster, with less friction. The cycle feeds back on itself.
The Cost of Validating Late
There's a financial mechanic that large tech companies often ignore until it’s too late: the cost of repositioning a brand that has already generated antibodies in the market is exponentially higher than the cost of building trust from the outset.
OpenAI isn’t at that point yet, but trajectory matters. Every dollar spent on podcasts, policy documents, and sophisticated lobbying spaces is a dollar not invested in solving the concrete issues that foster disapproval. Unlike technical infrastructure, which generates depreciable assets, spending on narrative has a very short shelf life when not backed by substantial changes in product or corporate behavior.
Analyst Brittany Ellich anticipates that anti-AI sentiment will likely worsen before it improves, and that recovery will come through practical honesty: openly acknowledging what doesn’t work and what the technology is actually beneficial for. This thesis holds empirical validity. Markets respond well to honest specificity and poorly to broad promises that cannot be upheld in everyday use.
What the industry has before it is not a branding issue; it’s an opportunity to redesign its proposition based on the variables it is currently ignoring because they seem costly to implement. Transparency in training data, compensation for creators, legal assurances for businesses adopting these tools—each of these variables has a real cost, but they also represent a market willing to pay for them. A market that is currently choosing not to adopt AI because no one in the sector has exercised the discipline to eliminate the frictions that matter instead of accruing capabilities that can no longer be distinguished from one another.
The leadership this situation demands is not that of publishing a 13-page document on the social contract. It’s about having the conviction to change the proposal variables before the market forces it to, and doing so based on concrete commitments validated with real users, not with lawmakers in a well-decorated office in Washington. Spending capital on narrative to defend a position that the product has yet to justify is not strategy; it’s simply time management before the inevitable adjustment.
