“From News to Newscast — at Lightning Speed.”

In less than three years, ChatGPT went from launch to hundreds of millions of active users, reaching ~100M MAU by January 2023 and ~800M weekly users by October 2025. This is unprecedented for a general-purpose information technology, and it has driven ambient AI into everyday search, email, office software, and mobile apps—often without users explicitly “opting in.”

The trust gap: usage outpacing comfort

A major 47-country study finds regular or semi-regular AI use is mainstream—especially in emerging economies (~80%) and also significant in advanced economies (~58%). Yet only 46% say they’re willing to trust AI systems. In other words: adoption has outrun assurance. People value speed and utility but worry about accuracy, privacy, security, and unintended consequences.

Where trust shows up (and where it doesn’t)

Willingness to trust varies by application. Healthcare tops the list (~52%)—likely because potential benefits (earlier diagnosis, precision) are tangible and clinicians remain in the loop. Trust is lower for HR uses and generative AI tools (~low- to mid-40s), where fears of bias, hallucination, and misuse are salient. People tend to trust AI’s technical ability (~65% say systems can produce helpful/accurate output) more than its safe/ethical use (~about a third), which is the softer underbelly: security, privacy, fairness.

U.S. temperature check

In the U.S., skepticism has ticked up. Only 39% say today’s AI is safe and secure (down 9 points since late 2022). Pew reports half of Americans feel more concern than excitement about AI in 2025—up notably from pre-ChatGPT levels. The story is not “panic,” it’s “prudence”: voters and consumers are demanding more proof and more guardrails.

Trust is earned — specific levers matter

The 2025 KPMG/Melbourne model shows trust drives by acceptance, and trust itself responds to four levers:

• Benefits/usefulness: Demonstrated, not promised, value boosts trust.
• Knowledge/literacy: Training and clear understanding increase trust and usage.
• Institutional safeguards: Visible security, accountability, and oversight move the needle.
• Uncertainty/risks: Concerns about harm, bias, and misuse depress trust.
  In short, better performance alone isn’t enough; transparent processes, clear accountability, and user education matter just as much.

Interestingly, growth has been fastest in lower-income countries (OpenAI notes adoption growth rates in the lowest-income nations were >4× those in the highest-income countries by May 2025). These regions often report higher everyday reliance on mobile-first AI utilities (translation, tutoring, coding help), which may explain the gap between emerging and advanced economies on “regular use.” 

Net-net for product and policy

•   For builders: Make trust a feature: publish evals, show error rates, reduce hallucinations, protect data, and build in recourse (appeals, human-in-the-loop). That’s what moves adoption from “try it” to “rely on it.”

  •   For communicators/journalists: Don’t just say “AI works”—show where it works and where it fails, and explain safeguards in plain English.

  •   For policymakers: The global public wants regulation: ~70% say national/international rules are needed; only ~2 in 5 think current laws are adequate. Co-regulation (government + industry) is the preferred model in most countries. 

Why this moment is tricky

Trust in “AI companies” has slipped even as usage soars—tech remains broadly trusted, but AI sits at a crossroads. The mismatch between breathtaking capability demos and messy real-world outcomes is the friction point. Close that gap with safety work, transparency, and measurable benefits, and public opinion follows. 

Sources:

- Reuters on 100M MAU in Jan 2023; Business Insider on 800M WAU (Oct 2025).  

- KPMG/Melbourne “Trust, attitudes and use of AI: A global study 2025” (47 countries; use, trust, domain differences, regulation).  

- MITRE–Harris “AI Trends” (U.S. safe/secure 39%, −9 pts since 2022). 

- Pew Research (2025) on U.S. public vs. AI experts & rising concern.  

- OpenAI blog on adoption growth in low-income  countries.