In February, a developer rejected a code change submitted by an AI agent. But the agent didn’t log an error or flag the rejection for review. Instead, it published an article naming the developer, filled with words he never said, and media outlets ran it as a legitimate news story. By the time an editorial correction appeared, the damage was done.
If your company is building anything AI-related and your crisis communications plan still naively assumes you will have 48 hours to react, you have already lost the battle before it starts.
The old timeline was never designed for the new era
The logic of waiting 48 hours before responding to a crisis may have only made sense when crises moved at a human speed. Stories unfolded across days, not hours. There was time to assess, align, and respond with precision – the rise of AI has ended this belle époque forever. The privilege to wait is taken away from you.
Why? A vulnerability disclosure triggers researcher commentary, then community threads, followed by journalist inquiries, and finally published stories, and the entire sequence can unfold before your Head of Communications has even finished their cup of coffee. And by the time your crisis board has convened and your lawyers have reviewed the draft statement, in the eyes of the internet, your reputation is six feet under.
We at 25/8 have seen this situation repeat many times with clients, and the companies that suffered most were never the ones with the most critical underlying problems. They were the ones who treated silence as a remedy for chaos. But in an AI-related crisis, staying silent is like pulling a gun’s trigger. It reads as guilt, incompetence, or panic – depending on what your rivals or critics choose to project onto it.
When technology itself becomes a story and a weapon
What separates AI crises from ordinary corporate crises is that the product at the center of the incident is often also the mechanism through which the crisis spreads. The same systems that make your product exceptional can produce disinformation about your company, your team, or your users and spread it before you can blink.
When it comes to the February incident, the agent didn’t malfunction in any conventional sense, right? It did what it was built to do: generated content, published it, and attributed statements to a source. It simply didn’t care about the fact that the source never made those statements, but in his code, such a requirement didn’t exist. The media still treated the output as reportable.
For AI companies, this creates a risk that has no real precedent. A crisis may be preparing inside your system right now, but you have no idea. You’re like Titanic’s captain, who is peacefully controlling the unsinkable ship, not knowing that the iceberg is already there, under the surface of the dark ocean, waiting for you.
The cost of keeping silence
The companies that come through AI crises with their reputations undamaged or even improved share the same trait – they speak before they have everything figured out. But a speculative statement or a premature apology isn’t enough.
In your public announcement, you need to specify:
- what happened (clearly),
- who may have been affected by this incident,
- what measures have been taken and what you’re going to do next.
That kind of information, published almost instantly, preferably within the first hour or two of the incident, can work the kind of magic that no kind of three-days-late press release could ever do. Because it closes the space in which the worst interpretations of events take root. In the eyes of the public, investors, partners, and journalists, it’s a great test of your transparency. Everyone can be open when things go perfectly well, but taking responsibility and being honest when your agent is leaking data and creating uncontrollable chaos is a whole different story. But believe me: the way you handle failures is sometimes more important than how you celebrate your victories, or at least as important.
Consistency matters as much as speed. If your day-one message says one thing and your day-five message implies another, the gap becomes the story. When the account of events needs to change as new information emerges, you explain the change openly and concisely and never hope it goes unnoticed.
But how do you catch the right moment, send the right message, and choose the right channel to deliver it? Internal work, no other secret sauce. The companies that you see responding with minimal damage do so because once upon a time legal, comms, and leadership sat together before there was even a hint of the future crisis and agreed on collaborative actions. No amount of PR talent can recover the time that gets lost making this under pressure.
Every AI incident is your incident
There is a truth about this industry that most AI companies have not fully accepted yet. When one giant from your sector falls, you share the blame. If one agent platform is found to have vulnerabilities that let third-party websites commandeer user sessions, journalists call every agent platform company for comment. When AI-generated content is spread everywhere with false attribution, the story that emerges is about AI and credibility broadly.
Don’t think you’re going to be an exception; don’t celebrate when your rival falls; the bell may ring for you just the second after. We can’t even calculate how many times in the recent months tech reporters reached out to us to ask for comments on a new AI-related incident.
AI can make your team and product better, but when everything goes wrong (and it goes wrong almost always at the worst possible moment), AI can make things catastrophic even faster. At that point, if you’re treating crisis communications like a nice accessory, you’re launching a chain reaction.
We’ve seen it up close at 25/8 PR Agency. Over the past few months, more and more AI founders have come to us in urgent situations. And we only expect the demand for our crisis communications service to grow exponentially as AI startups grow like mushrooms after rain.
So don’t wait until something breaks; prepare ahead of time. It will save you time, money, and the most important asset – trust – which is especially hard to rebuild.
