If you're using AI in your marketing operation and haven't had a serious conversation about governance, risk, and brand protection yet, you're not being innovative. You're being negligent.
SXSW 2026 sent a crystal-clear message: AI isn't a playground; it's a minefield. And most brands are marching through it blindfolded.
Michelle Lee, former VP of AI at Amazon and White House advisor, didn't mince words. She presented a framework of the 7 risks of AI that every C-level executive should have printed on their office wall. And the case she brought to illustrate the danger is chilling: a finance executive transferred $26 MILLION to fraudsters after being fooled by a deepfake video conference with his CFO.
This isn't a Black Mirror episode. It actually happened.
In this article, On is going to dissect the 7 deadly sins of AI in marketing and build a practical playbook so you can innovate without blowing up your own reputation.
The 7 Deadly Sins of AI (and How to Avoid Them)
The euphoria around generative AI made us forget a basic rule: every powerful tool comes with an instruction manual and a list of warnings. Ignoring them is a recipe for disaster.
Sin 1: Hallucination (The Confident Lie)
AI doesn't "know" anything. It's a probability machine. It delivers the most likely answer, not necessarily the correct one. The danger is that it does so with a self-confidence that would make a self-help guru blush.
The Risk to the Brand: Using AI-generated data without verification for reports, posts, or campaigns can destroy your credibility in seconds. Imagine publishing a "hallucinated" market data point that turns out to be false.
The Penance: Triple Validation. No AI-generated information should be published without being checked against at least two reliable primary sources. Implement a fact-checking process for all AI-assisted content.
Sin 2: Misinformation and Deepfakes (The Reality Crisis)
The $26 million case is just the tip of the iceberg. Audio, video, and image deepfakes are becoming indistinguishable from reality. Your brand could be targeted by attacks that use your CEO's image to announce a fake recall, or a video of a "real" customer complaining about a product they never used.
The Risk to the Brand: Total erosion of trust. If your audience can no longer tell the difference between official communication and a hoax, your brand loses control of the narrative.
The Penance: Digital Watermarking and Encryption. Adopt standards like C2PA (Coalition for Content Provenance and Authenticity) to add a cryptographic verification layer to your official content. Educate your audience to look for these authenticity seals.
Sin 3: Bias (Silent Discrimination)
AI is a mirror of the data it's trained on. And our data is inherently biased. The example from SXSW was brutal: a US healthcare system began prioritizing white patients over Black patients for certain treatments. The reason? The algorithm wasn't trained on race data, but on historical medical spending data, which, for complex socioeconomic reasons, correlates with race.
The Risk to the Brand: Launching a campaign or product that unintentionally discriminates against or offends a demographic group. The reputational damage can be irreparable.
The Penance: Bias Audits and Red Teaming. Before launching any AI system that impacts customers, form a red team with people from different backgrounds to try to "break" the algorithm and find its blind spots. Use bias-auditing tools to analyze training datasets.
Sin 4: Legal Uncertainty (The Copyright Roulette)
Is your company using Midjourney or DALL-E to create images? Congratulations, you're in a legal minefield. Can AI be the author of a work? The rulings from US courts and the USPTO (the patent office) are a resounding NO. If the creation was made with a simple prompt, it has no copyright protection. If it was an iterative process, with a lot of human intervention, it might.
The Risk to the Brand: Creating an entire visual identity for a client that, legally, can't be protected and can be copied by anyone. Or, worse, using a generated image that infringes an artist's copyright without knowing it.
The Penance: A Clear Intellectual Property Policy. Set internal rules on the use of generative AI for client work. Document EVERY step of human intervention in the creative process. Prioritize AI platforms that use only licensed image banks and offer legal indemnification, like Adobe Firefly.
Sin 5: Confidentiality (The Glass Wall)
An employee at a major tech company copied and pasted notes from a confidential internal meeting into a public chatbot to "summarize" them. The result? Company secrets indexed and available to the world. "Enterprise" versions of chatbots can protect against external leaks, but not against internal naivety (or malice).
The Risk to the Brand: Leaks of strategies, customer data, unreleased product information... the list is endless.
The Penance: Training and Sandboxing. Educate your team on what can and can't be entered into public AI tools. Create sandboxes (testing environments) with private LLMs, running on-premise or in virtual private clouds (VPCs), to work with sensitive data.
Sin 6: Regulation (The Inevitable Bureaucracy)
The EU AI Act, the world's first major AI regulation, is already a reality. It classifies AI applications into 4 risk levels, from "unacceptable" (banned) to "minimal" (unrestricted). Other countries will follow the same path. Ignoring this is like building a factory without looking at environmental laws.
The Risk to the Brand: Heavy fines and the need to completely redesign systems and processes that weren't built in compliance with the law.
The Penance: Monitor and Adapt. Designate someone responsible for tracking the evolution of AI regulation in Brazil and in the markets where your company and clients operate. Build a compliance framework that already incorporates EU AI Act principles, such as transparency and explainability.
Sin 7: Security (The Digital Backdoor)
Every new AI tool you integrate into your operation is a new entry point for cyberattacks. The same AI that helps personalize emails can be exploited to create ultra-convincing phishing attacks against your own team.
The Risk to the Brand: Leaks of customer data, ransomware attacks, and the loss of trust that underpins any business.
The Penance: Layered Security. Adopt a "Zero Trust" approach. Invest in cybersecurity training focused on AI threats. And above all, remember Michelle Lee's 70/30 rule: 70% of the success (and security) of any digital transformation depends on people and processes, and only 30% on technology.
The Golden Rule: 70/30
If there's one number to burn into your memory from this article, it's this: 70/30. Seventy percent of the success of any digital transformation depends on change management, people, and processes. Only thirty percent depends on the technology itself.
"Data is an organizational asset, not a departmental one." — Michelle Lee, former VP of AI at Amazon
This means investing millions in AI tools without investing in your team's culture, training, and governance is like buying a Formula 1 car and putting someone without a driver's license behind the wheel. Technology is useless without the right people, with the right processes, making the right decisions.
Practical Playbook: The AI Fortress for Your Company
- Create an AI Ethics Committee: A multidisciplinary group to evaluate new tools and projects through the lens of the 7 risks.
- Develop an “AI Constitution”: A living document that establishes the agency's rules of engagement with the technology.
- Invest in “Walled Garden” Tools: Prioritize AI platforms that operate in closed, secure environments (such as the Enterprise versions from Microsoft, Google, and Adobe).
- Document Everything: Create an audit trail for all content generated with AI assistance, especially visual content.
- Ongoing Education: Turn training on AI risks into a continuous process, not an annual workshop.
Innovating Responsibly
AI is, without a doubt, the most transformative technology of our generation. But with great power comes great responsibility. The brands that will thrive won't be the ones diving in headfirst without considering the consequences, but those that build a solid foundation of governance, ethics, and security.
For companies in general, it's not just about using AI, but mastering it. And mastering a tool means knowing its limits and its dangers as well as its potential.
Transformation depends on us. And so does responsibility.
Here at On, we turn SXSW's global trends and warnings into digital marketing strategies that truly generate measurable, safe results for your revenue.
So, if you want to understand how to implement solid AI governance in your business and protect your company, talk to us and keep following our blog to lead the future responsibly.
