The Future of Trust and Distrust in an AI World Webinar

Overview

Roy Morgan Risk Labs’ June 2026 quarterly update focuses on how trust turns into distrust in an AI-driven world. It argues that the biggest failure of brand leaders is measuring the “good news” while missing the signals that customers are actively preparing to leave.

Key arguments and findings

• Distrust behaves like an “early warning” risk state, not a simple drop in reputation. Trust is framed as an earned expectation, while distrust is active suspicion—the belief that a company will put itself first when it matters. Distrust spreads faster than indifference and leads to customer loss and escalation.

• Existing brand metrics miss distrust because they don’t ask about risk. Reputation/NPS/satisfaction-style instruments are described as measuring positive standing (e.g., leadership, innovation, governance), but not whether people are becoming angry, suspicious, or ready to reject. The update also claims that when people explain trust/distrust, they rarely cite reputation: only 9% of people who trust and 6% of those who distrust reference “reputation” as the reason.

• Australians increasingly distrust—especially after 2020. Using Roy Morgan’s risk monitoring since 2017, the update says Australia shifted from “trusting” pre-mid-2020 to sustained distrust. The nadir cited is March last year, when 70% distrusted at least one company. Latest figures cited still show 65% distrusting vs 47% trusting.

• Trust changes are usually “stable at the top,” while distrust dynamics drive movement. The most trusted brands are said to show little churn (including Bunnings, Aldi, Kmart, CBA, and Apple). Meanwhile, ranking movement is often attributed to distrust falling rather than trust rising.

  “You can’t improve trust until you get distrust under control.”

• Top “most distrusted” brands remain unchanged, but new entrants reveal how distrust accelerates. The most distrusted top five cited as stable are: Optus, Meta, Telstra, Woolworths, and Coles (with Optus noted as #1 since January; Telstra’s mention is linked to a later outage). New entries include Hancong Prospecting entering the top 20 for the first time, alongside rising negative sentiment toward Gina Reinhardt.

• Distrust can spread without familiarity—through narratives and virality. Palantir is cited as entering the top 30 most distrusted despite many Australians reportedly not knowing the company a year earlier. The update argues distrust travels digitally and can become viral through campaigns and association—not necessarily via direct company failure.

Case studies: how distrust “converts” into scandal

Coles (supermarket arc)

• Distrust built gradually through the cost-of-living crisis, then plateaued/recovered due to Coles’ response efforts.
• A federal court proceeding and subsequent findings are described as a turning point—especially after activism linked Coles to Palantir.
• The update claims that the reasons people give for distrusting Coles shifted from pricing complaints toward dishonesty, and that one legal finding “undoes a year of trust-building.”

Woolworths “natural experiment”

• Woolworths is described as showing a similar distrust pattern, but the update suggests its chart didn’t worsen in this quarter likely because its federal court case hadn’t been decided yet.
• The implication is that distrust is “waiting for the trigger.”

AI connection: why “adoption” is not “trust”

• AI brands are rising on distrust charts even without a single defining “disaster.” The update claims multiple AI brands are among the most distrusted: OpenAI (ChatGPT) in the top 20 and Palantir at ~30, with Google also drifting negatively.

• Distrust is attributed to ethics/privacy concerns and perceived dishonesty—not performance. OpenAI’s distrust trend is described as soaring while trust stays relatively flat. Respondents cite that it doesn’t protect privacy, is unethical, and is dishonest. These concerns are framed as anticipatory objections—meaning negative interpretation can form before any formal event occurs.

• A “risk register” is named by ordinary people. When asked what worries them about AI, respondents allegedly list risks unprompted, including: loss of ability to think, inaccuracy, disinformation/deepfakes/scams, jobs, privacy/surveillance, environment, and loss of control. This is framed as an early warning map that companies/consultants may fail to capture.

Proposed solution: “Risk intelligence operating system”

The update’s core idea is that boards and leaders need measurement that detects where rejection is forming, not just whether customers are satisfied.

• Invert survey questions from “positive outcomes” to explicit rejection/distrust measurement:

  • Add measures like “Would you reject us and why?”
  • Ask what drives dissatisfaction and what made someone distrust—not just whether they’d recommend.

• Behavioral insight is missing from positive-only scoring. The update argues that harmful behaviors (refusing to return, organizing boycotts, agitating for punishment) are not just low scores—they’re a different phenomenon, and positive-only metrics don’t capture them at all.

• Risk intelligence is the gap-finder between machine-speed capability and human-speed permission. Central thesis: AI accelerates capability, but trust/distrust changes more slowly, creating the danger that real-world consequences can outpace governance before people feel safe.

Mentioned “watch list” going forward

The update urges monitoring:

• Telstra
• Woolworths pending court judgment
• AI brands’ trust/distrust trajectories

Presenters / Contributors

• Michelle Lavine (Roy Morgan Risk Labs)
• Roy Morgan Risk Lab (organization; no other individual presenters named in the subtitles)