Last month, Anthropic came under fire as its new Claude Fable 5 was abruptly taken out of action due to significant security fears, with the US government ordering its suspension just three days after launch. The model had fallen victim to jailbreaking, whereby hackers get past restrictions there to protect a cyber network, and are able to then access sensitive information. It took 19 days, a rebuilt set of safety classifiers and a series of new commitments to the US government before the model was restored to service at the start of July.
This is one of the most commonly used AI programmes around and the episode has inevitably created a few jitters across the industry. At the same time, thousands of technology companies are launching AI solutions or integrating AI into existing systems and episodes like Claude’s don’t help with building trust. Notably, what it took to get the model back online – new safety classifiers, a proposed industry framework for rating the severity of jailbreaks and closer collaboration with government – says a lot about what rebuilding that trust now requires.
For companies launching AI-related products, getting cut-through in the media gets harder. In other words, while there is no shortage of interest in AI, there is also no shortage of scepticism. After two years of relentless AI announcements, journalists have become increasingly selective about which companies they cover and which claims they take seriously.
As a result, a new credibility filter is emerging. Journalists know that readers increasingly care about how AI systems are built, not just what they can do.
This shift has created what we are calling the ‘AI trust-ometer’. These are five areas where we consistently hear questions from journalists, looking to establish legitimacy, before they’ll cover an AI product in their editorial.
Question 1: Where Does the Data Come From?
In the past, few technology journalists would have asked detailed questions about training data. Today, it is often one of the first issues raised. Reporters want to understand what data was used to train the model. Was it licensed? Was consent obtained? Is customer data used for training? And crucially, how is sensitive information protected?
High-profile lawsuits, copyright disputes and concerns over data privacy have pushed data provenance into the mainstream. As a result, companies that can clearly explain their data practices are often viewed as more credible than those that rely on vague assurances or technical jargon.
Question 2: Who Is Governing the Technology?
One of the most significant changes in AI communications is the rise of governance as a media topic. Governance used to sit quietly within legal and compliance departments, now it is increasingly part of the public narrative. Journalists want to understand who oversees AI development and what safeguards are in place? Is there a responsible AI framework and who is accountable when things go wrong?
What this demonstrates is that increasingly AI is seen as a technology capable of influencing decisions, behaviours and outcomes at scale. That means governance is becoming a trust signal.
Question 3: How Do You Manage Risk?
Every AI company can describe the opportunities its technology creates, but how many can clearly explain the risks? This is where many media interviews can start to unravel.
Reporters are increasingly asking how you test model performance. Prevent hallucinations? What human oversight exists? The answers really matter because journalists will not always be looking for a positive news story. Being transparent can help with creating that trust that we’re looking for.
Question 4: What Standards or Regulations Do You Meet?
The AI regulatory landscape is evolving rapidly and journalists are increasingly interested in whether companies are building compliance into their products. They may want to know how you are preparing for emerging regulations. What industry standards do you follow and how do you address privacy requirements?
Companies that view regulation as an opportunity to demonstrate maturity can use regulatory readiness as a strong differentiator.
Question 5: Can You Prove It Delivers Value?
Only after trust has been established do journalists return to the question many AI companies want to answer first which is does the product actually work.
The challenge is that every AI platform promises transformation or productivity gains. Every startup says it is revolutionising an industry and reporters want evidence. They want measurable business impact and independent validation, simple and tangible specifics that will build credibility.
In the past marketing teams could lead with vision, but now with such a saturated market, credibility often precedes innovation. Before journalists are willing to write about what an AI company can do, they increasingly want reassurance about how it was built, how it is governed, how risks are managed, and whether the claims being made can be substantiated.
Trust is the product that must be sold first
Photo Credit: istock elenabs
