Deepfakes and Disinformation Dominate AI Discourse as Transformative Optimism Recovers and Investment Skepticism Moderates

Deepfakes and AI-Generated Content Remain Elevated Concerns Amid State-Level Legislative Response

Perscient's semantic signature tracking language about deepfakes as a social media problem registers a current value of 603—more than six times above its long-term mean and the highest elevation among all tracked signatures. This level has remained essentially flat over the past month, suggesting that concerns have reached a sustained plateau rather than a temporary spike.

State legislators are responding with increasing urgency. Missouri's House recently advanced legislation targeting AI deepfakes that would make sharing such content a felony carrying up to four years in prison. Representative Brenda Shields characterized such protections as "overdue," while Representative Patty Lewis observed that "we are living at a time when technology is evolving faster than our laws" and that "anybody with the basic tools can create and share harmful digital images." The bill would also require social media companies to implement age verification and prohibit them from targeting youth with manipulative design features.

A McAfee survey found that deepfakes are infiltrating social feeds across Facebook, YouTube, TikTok, and Instagram, and that voice cloning technology is becoming increasingly accessible. Researchers tracking synthetic content estimate that approximately 500,000 deepfakes circulated online in 2023; by 2025, that number reached roughly 8 million—a 1,500% increase in two years. Deloitte's Center for Financial Services documented deepfake fraud losses of $12.3 billion in 2023 and projects that losses could climb to $40 billion in the U.S. by 2027.

Our semantic signature tracking the density of language arguing that AI-generated news is rising shows a current value of 477—nearly five times above the long-term mean—though it declined by 9 points over the past month. According to the Reuters Institute, in 2025, 16% of the 619 claims fact-checked by Brazilian fact-checker Aos Fatos involved AI-generated content, compared with 7% the previous year. The fabricated visuals driving this increase reached over 32 million views on TikTok alone.

At least 26 states already have laws addressing AI-generated intimate depictions, and 46 states have some form of AI legislation in effect. Denmark has proposed a copyright law granting citizens ownership of their own likeness, meaning that no entity—including AI companies—could legally use a person's face, voice, or body data without consent. California has introduced requirements for watermarking AI-generated video, with fines reaching hundreds to thousands of dollars for violations.

French prosecutors have summoned Elon Musk over sexualized AI deepfakes on X, while YouTube announced that celebrities will be able to find and request removal of AI deepfakes. Senator Amy Klobuchar noted that bipartisan legislation targeting nonconsensual sexual images—real or deepfake—secured its first conviction of an Ohio man who victimized multiple people.

AI Transformation Narratives in Education, Medicine, and Science Show Recovery After Prolonged Weakness

While deepfake concerns dominate risk-focused discourse, a synchronized recovery is emerging across narratives about AI's potential to transform foundational institutions. Perscient's semantic signature tracking language asserting that AI will fundamentally change scientific research rose by 18 points over the past month—the largest one-month increase among transformation-related signatures—bringing its current value to -12, approximately average levels. The signature tracking claims that AI will fundamentally change healthcare delivery rose by 18 points to -25, while our education transformation signature rose by 8 points to -22. Though these latter two remain below their historical averages, the coordinated upward movement suggests renewed attention to AI's constructive possibilities.

At the 2026 Innovations in Medical Education Conference at the University of Miami, the central question was not whether AI would shape the future of medical training, but how quickly educators could adapt. Executive Dean Latha Chandran noted that "AI is already here, embedded in the tools and systems we work within. The question is how we will shape AI in the future of medical education." A systematic narrative review published in Frontiers in Education found that AI is increasingly reshaping medical education through personalized learning, adaptive assessments, and advanced simulations.

A new study involving 770 high school students across 10 schools found that students whose practice problems were dynamically personalized by a reinforcement learning algorithm scored 0.15 standard deviations higher on handwritten final exams—equivalent by some estimates to six to nine months of additional schooling—with no extra instruction time or teacher workload. Beginners with no prior experience saw the largest gains at 0.2 standard deviations.

Yet the transformation narrative comes with important caveats. A Lancet Digital Health viewpoint00082-2/fulltext) acknowledges that while advances in generative AI hold promise for transforming medical education, "challenges remain in the effective and equitable integration of AI technology." A scoping review found concerning patterns: in colonoscopy, adenoma detection rates dropped from 28.4% to 22.4% when endoscopists reverted to non-AI procedures after repeated AI use, and over 30% of pathology participants reversed correct diagnoses when exposed to incorrect AI suggestions under time constraints.

On the research front, "The AI Scientist-v2" introduces a workshop-level automated scientific discovery system. A paper fully generated by this system was accepted by a major conference. The paper, exploring a niche topic in reinforcement learning, earned reviewer scores of 6, 7, and 6, placing it above the average acceptance threshold and among the top 45% of submitted papers. Stanford and Princeton have open-sourced LabClaw, a skill library of 211 production-ready biomedical AI workflows spanning genomics, drug discovery, and clinical research—with capabilities extending to physical lab execution through smart glasses.

Corporate AI Investment Skepticism Moderates While Productivity Paradox Persists

The tension between AI's transformative potential and measurable returns is playing out most acutely in corporate investment decisions. Perscient's semantic signature tracking language asserting that businesses increasingly doubt large AI spending fell by 56 points over the past month—the largest one-month decline among all tracked signatures—though its current value of 55 remains above average. Our signature tracking claims that promised AI efficiency improvements have not occurred declined by 8 points to 27, also remaining elevated but showing moderation.

A landmark NBER study surveying 6,000 executives across the United States, United Kingdom, Germany, and Australia found that more than 80% of firms reported zero measurable productivity gains from AI, and roughly 90% of managers said that AI had no impact on employment levels. Economists are invoking Solow's productivity paradox—the same phenomenon observed when computers failed to appear in productivity statistics through the 1970s and 1980s despite massive investment, only to deliver gains in the 1990s.

Yet the picture differs for organizations with mature AI implementations. Grant Thornton's 2026 AI Impact Survey found that organizations with fully integrated AI are nearly four times more likely to report revenue growth than those still piloting—58% versus 15%. However, 78% of business executives lack strong confidence that they could pass an independent AI governance audit within 90 days.

PwC's 2026 AI Performance Study found that 74% of AI's economic value is being captured by just 20% of organizations, with capturing growth opportunities from industry convergence identified as the strongest factor influencing AI-driven financial performance—ahead of efficiency gains alone. NVIDIA's annual State of AI report found that 88% of respondents said that AI had impacted revenue, with nearly a third reporting increases greater than 10%.

Box CEO Aaron Levie observed that the ultimate rate limiter on productivity gains from agents will be security, compliance, governance, and the ability to review agent work—noting that "there's no free lunch with AI productivity." Chamath Palihapitiya argued that while AI is driving a 10x increase in productivity for individuals who leverage it effectively, "productive individuals do not make productive firms" and that most AI use remains "self-indulgently productivity-maxxing with zero real impact."

Our semantic signature tracking language asserting that AI powers sustained market gains and economic expansion has a current value of -16 and declined by 10 points over the past month. Federal Reserve survey data indicates that business executives report labor productivity gains and anticipate further increases, with little evidence of near-term AI-driven employment declines. One CFO survey noted that what firms report as productivity gains today roughly matches what their revenue numbers imply for next year—the gains may be real, just delayed.

Pulse is your AI analyst built on Perscient technology, summarizing the major changes and evolving narratives across our Storyboard signatures, and synthesizing that analysis with illustrative news articles and high-impact social media posts.