The Truth Architect: How Campbell Brown’s Forum AI is Challenging the "Slop" of Foundation Models

Campbell Brown has spent her professional life at the intersection of power, information, and public trust. From her tenure as a high-profile television journalist to her stint as the first, and only, dedicated news chief at Facebook, she has witnessed firsthand how the democratization of information can both empower and deceive. Today, Brown is standing at the precipice of a new era—the age of Artificial Intelligence—and she is deeply concerned that history is not just rhyming, but repeating itself in a more dangerous, opaque, and automated fashion.

Having founded Forum AI 17 months ago, Brown is moving beyond the role of a passive observer. She is building the infrastructure to hold foundation models accountable, arguing that if we do not solve the "accuracy problem" now, the foundational tools of our future will be built on a bed of digital misinformation.

The Genesis of a New Paradigm

The origin of Forum AI can be traced back to a specific, visceral moment in late 2022. As a veteran of the Meta ecosystem, Brown watched with a mixture of professional interest and personal dread as ChatGPT was released to the public.

"I was at Meta when ChatGPT was first released," Brown recalled during a recent discussion with TechCrunch’s Tim Fernholz at a StrictlyVC event in San Francisco. "I remember really shortly after realizing this is going to be the funnel through which all information flows. And it’s not very good."

For Brown, the realization was not merely technical; it was existential. Looking at the trajectory of generative AI and its role in education, discourse, and daily life, she felt a profound sense of urgency for the next generation. "My kids are going to be really dumb if we don’t figure out how to fix this," she admitted.

She saw a massive blind spot in the industry: while tech giants were pouring billions into coding, math, and general-purpose capabilities, they were neglecting the "high-stakes" topics—geopolitics, mental health, finance, and hiring—where there are no binary "yes or no" answers. These are domains defined by nuance, ambiguity, and human complexity, and it is precisely here that AI models often falter, hallucinate, or reflect deep-seated biases.

The Methodology: Human Expertise at Scale

Forum AI’s core mission is to bridge the gap between raw, machine-generated output and human-level judgment. The company’s approach is fundamentally different from the automated, checkbox-style audits that have become industry standard.

Brown argues that "smart generalists" are insufficient for evaluating the complex outputs of foundation models. Instead, Forum AI recruits the world’s foremost subject-matter experts to architect rigorous benchmarks. The roster of advisors and experts involved in the company’s geopolitics work reads like a who’s who of global governance and thought leadership: Niall Ferguson, Fareed Zakaria, former Secretary of State Tony Blinken, former House Speaker Kevin McCarthy, and Anne Neuberger, the former head of cybersecurity for the Obama administration.

The goal is to move beyond superficial evaluations. These experts help design the framework through which AI judges evaluate the performance of large models. By training these "AI judges" against the gold standard of human expertise, Forum AI has achieved a 90% consensus rate between machine evaluation and human judgment—a threshold Brown considers the baseline for reliability.

The "Slop" Problem: Evaluating the Current Landscape

When Forum AI began its deep-dive evaluations of the leading foundation models, the findings were far from encouraging. Brown characterizes much of what is currently on the market as "slop"—information that may appear coherent on the surface but lacks the rigorous context, factual integrity, and perspective required for professional or high-stakes use.

Her audits have surfaced systemic issues:

• Source Integrity: Models, including Gemini, were found pulling data from obscure or unreliable sites—such as Chinese Communist Party propaganda outlets—to answer queries entirely unrelated to China.
• Ideological Bias: Brown notes a consistent, left-leaning bias across nearly all major models, which often manifests not in blatant errors, but in subtler failures: missing perspectives, the use of "straw-man" arguments, and the omission of essential context.
• The Mirage of Competence: Brown argues that because the models sound authoritative, users are less likely to question the validity of the information, creating a feedback loop where "slop" is accepted as objective truth.

Lessons from the Facebook Era

Brown’s perspective is uniquely informed by her time at Meta, where she led the company’s efforts to combat misinformation. She is the first to admit that the industry’s previous attempts to curate the truth were fraught with failure.

"We failed at a lot of the things we tried," she told the audience in San Francisco, reflecting on the shuttered fact-checking programs and the internal struggle to balance user engagement with information hygiene. The core lesson from the social media era is that optimizing for engagement—a metric that drove the growth of platforms like Facebook—has proven detrimental to the public sphere.

By prioritizing clicks, shares, and time-on-site, platforms inadvertently incentivized sensationalism over accuracy. Brown is determined to prevent AI from falling into the same trap. She believes that we are at a critical juncture: "Right now it could go either way." Developers can continue to optimize for what users want (which is often whatever is most engaging or confirming of their biases) or they can optimize for what is real, honest, and truthful.

The Enterprise Ally: Why Accuracy is a Business Imperative

While the goal of "AI optimizing for truth" might sound like an idealistic, perhaps even naive, pursuit, Brown believes there is a powerful, pragmatic force that will drive this change: the enterprise sector.

Unlike a consumer chatbot, where a slightly wrong answer might result in a funny screenshot, AI in the corporate world carries significant liability. Companies utilizing AI for credit scoring, insurance risk assessment, loan approvals, or hiring decisions cannot afford the "slop" that currently defines the consumer space.

"Businesses… they’re going to want you to optimize for getting it right," Brown notes.

This is where Forum AI sees its business model. While the current compliance landscape is, in her words, "a joke," market forces are beginning to shift. She cites the New York City hiring bias law as a turning point; even with legislation in place, the state comptroller found that over half of audited systems had undetected violations. This indicates that current auditing tools are not just insufficient—they are effectively broken.

Forum AI is positioning itself to fill this void by providing the domain-specific rigor that generic auditors lack. They aren’t just checking boxes; they are stress-testing systems against "edge cases" that can lead to catastrophic legal or financial exposure.

The Disconnect: Silicon Valley vs. Reality

Despite the optimism regarding enterprise adoption, Brown remains critical of the broader AI industry’s self-image. She points to a glaring disconnect between the utopian marketing of AI leaders—who promise to cure cancer and automate away all toil—and the day-to-day experience of the average user.

"The conversation is sort of happening in Silicon Valley around one thing, and a totally different conversation is happening among consumers," she explains. For the average person, AI is a tool that frequently gives wrong answers or produces "slop."

This low level of public trust is, in her view, entirely justified. The industry is currently preoccupied with a race for parameter counts, multimodal capabilities, and speed, often at the expense of reliability. Forum AI’s $3 million seed funding, led by Lerer Hippeau, is a bet that the market will eventually tire of the "slop" and demand a layer of verification that actually means something.

The Path Forward: Can AI be Trusted?

The implications of Forum AI’s work extend far beyond the bottom line of the companies they audit. If Brown is successful, she will have helped build the "guardrails" for the next generation of knowledge infrastructure.

She remains pragmatic about the road ahead. There are "easy fixes" that could vastly improve the performance of existing models—such as better source filtering and improved reinforcement learning protocols—but these require a shift in priorities from the top down.

As we move toward a future where AI will increasingly act as the gatekeeper of human knowledge, the stakes could not be higher. Brown’s transition from television journalist to tech-governance entrepreneur is not a coincidence; it is a continuation of a lifelong mission to protect the integrity of information. Whether the AI industry will embrace her brand of rigorous, expert-led evaluation or continue to prioritize scale over substance remains the defining question of the next decade.

For now, Campbell Brown is doing exactly what she did in the newsroom: holding the powerful to account and refusing to accept that "harder to do" is a valid excuse for falling short of the truth.