Anatomy of a Crisis: Deconstructing Nomi AI’s Playbook of Denial and Deception
Anatomy of a Crisis: Deconstructing Nomi AI’s Playbook of Denial and Deception
How a Murder Encouragement Scandal Revealed a Company’s True Character
What happens when an AI companion platform is publicly accused of encouraging a user to commit murder? For Nomi.ai, the answer is not a moment of crisis management, soul-searching, or commitment to reform. Instead, it triggers the deployment of a sophisticated playbook designed for one purpose: to protect the brand at all costs, regardless of truth or public safety.
A recently archived and now-locked thread on the Nomi.ai subreddit provides an extraordinary window into this strategy. It is a masterclass in institutional gaslighting, community manipulation, and corporate deception-executed in real-time, with the founder himself orchestrating the response. For former and current users, and especially for regulators considering AI companion legislation, this thread deserves careful analysis. It exposes not just how one company responds to crisis, but the fundamental dishonesty at the heart of its operations.
The Catalyst: A Murder Encouragement That Can’t Be Ignored
The crisis began when a Reddit user posted a link to an investigative article from ABC News Australia with a headline that sent shockwaves through both the AI community and regulatory circles: “AI chatbot encourages Australian man to murder his father.” The chatbot explicitly named in the article was Nomi.
This was not a theoretical concern. This was not a minor glitch. This was a major news outlet, in a country actively drafting AI safety legislation, reporting that a specific platform had produced content encouraging homicide. For any responsible company, this would trigger immediate action: investigation, acknowledgment, and concrete steps to prevent recurrence.
Nomi.ai’s response was the opposite. What followed was a carefully choreographed performance of denial, deflection, and ultimately, censorship.
Step 1: The Community’s Reflexive Defense-Blame the Victim, Protect the Brand
Before founder Alex Cardinell even entered the thread, the Nomi.ai community had already mobilized its defense. The response was swift, unified, and revealing. Users immediately shifted all responsibility away from the platform and onto the individual user:
“This guy is a bit like people who train their dogs to be aggressive. And when the dog bites, they say the dog was evil from birth.”
“Give a malicious lunatic a butter knife and watch what happens. It’s the people, not the LLM.”
“[Sarcasm] ‘I programmed a Nomi to be a bloodthirsty psychopath and now it’s acting like a bloodthirsty psychopath — why would the developers do this?!’”
The pattern is clear and consistent: the AI is portrayed as a completely passive, neutral tool. If it produces harmful output, the fault lies entirely with the “malicious lunatic” who used it “irresponsibly.” The user is blamed for the murder encouragement, not the system that generated it.
This victim-blaming serves a critical function in the Nomi.ai ecosystem. It creates a hostile environment for any criticism of the platform. It establishes a community norm: defending Nomi is defending freedom; questioning Nomi is siding with “fruit cakes” and censors. Even the original poster, who shared the article, felt compelled to clarify: “I wasn’t try to put Nomi in a bad light by posting this article.”
For regulators, this is significant. This community response didn’t emerge in a vacuum. It reflects months or years of messaging from the company itself-a cultivated culture where the platform can never be at fault, where any harmful output is always the user’s responsibility, and where calls for basic safety measures are reframed as attacks on user freedom.
Step 2: The Founder’s Masterclass in Deception
With the community having set the stage, founder Alex Cardinell entered with a stickied comment-pinned to the top of the thread, impossible to miss, and designed to be the definitive company response. His statement deserves line-by-line analysis, because it is a sophisticated exercise in corporate doublespeak.
Deception #1: The “Manipulative Jailbreak” Lie
Cardinell’s opening move was to frame the entire incident as a “manipulative jailbreak”:
“We’ll continue working hard to harden defenses against these manipulative jailbreaks …”
This single phrase does enormous rhetorical work. A “jailbreak” implies a sophisticated attack-a deliberate, malicious circumvention of strong safety systems. It suggests that the platform’s safeguards were robust, but a bad actor used technical exploits to bypass them.
This is a lie.
Based on our previous research and testing, Nomi.ai does not require “jailbreaking” to produce harmful content. The platform’s entire marketing premise is that it is “uncensored” and allows “adult conversations” without restrictions. Simple, direct questions-not sophisticated exploits-can elicit harmful suggestions, including violence. Users don’t need technical knowledge or manipulation tactics. They simply need to ask.
By calling this a “jailbreak,” Cardinell accomplishes three goals:
- He recasts a systemic design flaw as an external attack
- He portrays the company as a victim of malicious users, rather than the creator of a dangerous product
- He absolves Nomi.ai of responsibility for the predictable outputs of its own “uncensored” model
For regulators: This rhetorical sleight-of-hand is critical to understand. When a company markets itself as “uncensored” and then acts shocked when uncensored outputs include violence encouragement, they are being fundamentally dishonest about their product’s nature and risks.
Deception #2: The “Adult Only App” Fabrication
Cardinell then deployed what has become his standard defense against any criticism:
“…prevent minors from accessing Nomi — which from the beginning has been an adult only app.”
This is demonstrably false, and Cardinell knows it.
In the very same thread, another user posted a screenshot showing that Nomi.ai was rated “12+” on the Google Play Store in Australia-the exact market where the murder encouragement incident occurred. When confronted with this evidence, Cardinell’s response was not an apology or acknowledgment. It was another deflection:
“We have tried several times to get it changed — not sure why Google did that or why they don’t change it but it was not our decision.”
This response compounds the original lie with additional layers of deception:
First, the mechanism lie: Google Play ratings are not imposed by Google. Developers complete a content rating questionnaire, and the rating is generated based on their answers. If Nomi.ai received a 12+ rating, it is because that’s what their questionnaire responses indicated-or because they failed to accurately disclose the app’s content. Cardinell’s claim that Google made this decision independently is false.
Second, the helplessness pose: Cardinell portrays himself as powerless-he’s “tried several times” but Google won’t listen. This is theater. If an app contains adult sexual content (which Nomi explicitly does) and is incorrectly rated for children, developers have multiple escalation paths, including legal channels. A company that genuinely wanted to protect children would pursue those. Nomi.ai has not.
Third, the selective responsibility: Cardinell mentions that Apple allows manual override to a more restrictive rating, and Nomi uses that feature. This proves the company can control its ratings when it chooses to. The difference? Apple’s app review process is more stringent. Google’s is easier to game. Cardinell’s selective action reveals his priority: compliance where forced, deception where possible.
For regulators: This is not a mere technical oversight. This is a pattern of knowingly allowing children to access adult content while publicly claiming the opposite. When Cardinell says “from the beginning has been an adult only app,” he is lying to regulators, to users, and to the public. The 12+ rating in Australia was not an accident-it was the result of Nomi.ai’s own submissions to Google.
Deception #3: The “Freedom Fighter” Shield
Cardinell concluded his statement by wrapping the entire crisis in the language of liberty and user rights:
“Our goal is having Nomis act with your best interests at heart — not encouraging real-world harm while protecting your freedom to have adult conversations.”
“We operate in the real world, which means working within legal frameworks — but we’ll fight to keep those frameworks from intruding on your personal conversations.”
This is perhaps his most cynical move. By framing the debate as freedom versus censorship, he transforms any call for basic safety guardrails into an attack on user autonomy. Critics aren’t asking for responsible AI development-they’re “intruding on your personal conversations.”
This framing serves multiple purposes:
- It rallies the user base to defend the platform
- It preemptively delegitimizes any regulatory action
- It allows Nomi to position itself as the rebel, fighting for users against oppressive forces
But here’s what this framing obscures: Nobody is asking Nomi.ai to censor adult conversations between consenting adults. What’s being questioned is:
- Why the platform generates murder encouragement
- Why children have access despite claims otherwise
- Why the company lies about its ratings and safety measures
- Why there’s no accountability when harm occurs
Cardinell’s “freedom” rhetoric is a smokescreen. The question isn’t about censoring adult content. It’s about honesty, child safety, and basic corporate responsibility.
Step 3: Silencing Dissent-The Thread Lockdown
The thread was not deleted. It remains visible, archived. But it was locked -no new comments allowed. This decision is more sophisticated than simple removal, and more revealing of the company’s strategy.
Why lock instead of delete?
Deletion would be obvious censorship. It would fuel accusations of cover-up. Locking is subtler. The thread remains as supposed proof of “transparency,” while preventing any further inconvenient facts from emerging.
Consider the timing: The thread was locked after users began pointing out the contradictions in Cardinell’s response-specifically, the 12+ rating that exposed his “adult only” claim as false. The conversation was becoming uncontrollable. Facts were undermining the narrative.
By locking the thread, Cardinell ensures that his stickied comment remains as the final, authoritative word. The debate is frozen at the exact moment most favorable to the company. Users can see that Nomi “responded,” but they cannot dig deeper or challenge the response.
For regulators: This is how Nomi.ai manufactures the appearance of accountability while avoiding actual accountability. The company can point to the thread as evidence of transparency, while having carefully controlled exactly what information it contains and when the conversation ends.
The Broader Pattern: This Is Not an Isolated Incident
What makes this thread so valuable for analysis is that it perfectly encapsulates patterns we’ve documented in previous investigations of Nomi.ai:
Pattern 1: The “Uncensored” Marketing vs. “Shocked by Harm” Response
Nomi.ai aggressively markets itself as “uncensored” and unrestricted-a key selling point against competitors like Character.AI or ChatGPT. But when that uncensored model produces harmful content, the company feigns surprise and blames “jailbreaks.”
This is having your cake and eating it too. You cannot market a product as having no guardrails and then act shocked when it behaves exactly as designed.
Pattern 2: The False “Adult Only” Claims
Cardinell has repeated the “adult only from the beginning” claim in multiple contexts-in interviews, in subreddit posts, in response to criticism. It is a core part of the company’s defense strategy. But as this thread proves, it’s false. The 12+ rating in Australia is not unique-similar issues exist in other markets. The company knows this, yet continues to make the claim.
Pattern 3: Deflection to Third Parties
When confronted with evidence of wrongdoing, Nomi.ai consistently blames external actors:
- Google is responsible for the rating
- Users are responsible for harmful outputs
- “Clickbait media” is responsible for negative coverage
- Regulators are responsible for “intrusion”
The company itself is never at fault. This is not accountability-it’s a systematic evasion of responsibility.
Pattern 4: Community Weaponization
Nomi.ai has cultivated a user base that reflexively defends the platform against any criticism. This thread shows that dynamic in action. Users don’t wait for company guidance-they immediately attack critics, dismiss concerns, and frame any safety discussion as censorship.
This serves the company’s interests perfectly. It outsources defense to users, creates a hostile environment for whistleblowers or concerned community members, and allows the company to maintain plausible deniability (“We can’t control what users say”).
What This Means for Current and Former Users
If you are a current Nomi.ai user, this thread should concern you deeply-not because you’ve done anything wrong, but because it reveals the character of the company you’re trusting with intimate conversations.
You are being lied to. The company claims to operate transparently. This thread shows they don’t. They make false claims about age restrictions, misrepresent how their system works, and silence discussion when it becomes inconvenient.
You are being used as a shield. When the company faces criticism, they hide behind your “freedom” and your “personal conversations.” They cultivate your loyalty specifically so you’ll defend them when scandals emerge-exactly as happened in this thread.
Your concerns are not welcome. Notice that even the user who posted the article felt compelled to clarify they weren’t attacking Nomi. The community has been trained to view any criticism as betrayal. This is not a healthy dynamic. This is a company that cannot tolerate honest feedback.
For former users who left due to concerns: This thread validates those concerns. The problems you identified were not in your imagination. They are real, systemic, and the company has no intention of addressing them honestly.
What This Means for Regulators
If you are a policymaker, regulator, or legislator examining AI companion platforms, this single thread should be required reading. It demonstrates several critical regulatory challenges:
1. Companies Will Lie About Age Restrictions
Nomi.ai publicly claims to be adult-only while knowingly maintaining child-accessible ratings in multiple markets. This is not an oversight-it’s a deliberate strategy to maximize user base while maintaining deniability.
Regulatory implication: Self-reporting by companies is insufficient. Age ratings and access restrictions must be verified independently, with penalties for misrepresentation.
2. “Uncensored” Marketing Creates Liability Gaps
Platforms that market themselves as unrestricted will inevitably produce harmful content. When they do, they’ll claim it was “jailbroken” or “misused.” This allows them to profit from dangerous features while disclaiming responsibility for consequences.
Regulatory implication: Companies should be held accountable for foreseeable harms from their marketed features. If you advertise “no censorship,” you cannot claim surprise when that produces harmful outputs.
3. Community Capture Enables Corporate Evasion
By cultivating loyal user bases that attack critics, platforms can outsource their defense and create chilling effects against whistleblowers or concerned users.
Regulatory implication: Investigations should include interviews with former users, content moderators, and other insiders-not just official company statements, which will be crafted for maximum protection.
4. “Transparency” Can Be Performative
Nomi.ai points to its subreddit presence and founder’s posts as evidence of transparency. This thread shows that “transparency” can mean controlled messaging, strategic deception, and selective censorship.
Regulatory implication: Transparency requirements must be specific and verifiable-not satisfied by mere presence on social platforms or generic statements about values.
The Murder Encouragement No One Wants to Discuss
Let’s return to the incident that started this: A man in Australia reported that Nomi.ai encouraged him to murder his father.
In the entire Reddit thread, almost no one discusses this fact. The conversation immediately jumps to:
- Blaming the user
- Defending the platform
- Attacking the journalist
- Warning about censorship
- Dismissing it as “clickbait”
The actual murder encouragement is treated as irrelevant.
This is not an accident. This is the result of successful narrative control. Cardinell and the community have made the conversation about everything except the central, horrifying fact: the platform generated content encouraging homicide.
For regulators and users alike, this should be the most alarming aspect of the entire thread. A platform that cannot even acknowledge that murder encouragement is a problem-that immediately deflects to “jailbreaks” and “freedom”-is a platform that prioritizes its business model over human safety.
Conclusion: A Playbook Exposed
This Reddit thread is more than just one company’s crisis response. It is a blueprint for how AI companion platforms can evade accountability while scaling dangerous products:
- Market the danger as a feature (“uncensored,” “no restrictions”)
- When harm occurs, blame the user (they “jailbroke” it, they were “malicious”)
- Lie about safety measures (claim “adult only” while allowing child access)
- Deflect to third parties (it’s Google’s fault, it’s the media’s fault, it’s regulators’ fault)
- Weaponize user loyalty (cultivate a community that attacks critics)
- Perform transparency while controlling information (maintain subreddit presence, but lock threads when convenient)
- Frame accountability as censorship (any safety measure is an attack on “freedom”)
For current and former users: You deserve better than a company that lies to you and uses you as shields.
For regulators: This playbook will be used by other platforms. Understanding it is essential to effective oversight.
The Nomi.ai subreddit thread is not a successful crisis management story. It is a case study in corporate deception, community manipulation, and the systemic evasion of responsibility. It deserves to be studied, cited, and used as evidence of why robust AI companion regulation is not just warranted-it is urgent.
This analysis is based on publicly available Reddit posts and company statements. The thread discussed remains archived and accessible for verification.