AI abuses and scandals: the patterns

Every powerful technology gets misused, and AI is no exception. But “AI is dangerous” is a headline, not an insight. The useful move is to notice that AI abuses fall into a small number of recurring patterns — and that each pattern carries a specific, practical lesson for the people who use AI and the people who invest in it.

This article is about those patterns, not about pillorying any named company. As always, we describe categories and documented behaviour, and leave verdicts about specific firms to courts and regulators.

Pattern 1 — synthetic identity: deepfakes and voice cloning

The most financially damaging abuses in 2026 are impersonation. Cloned voices and deepfake video have been used to authorise fraudulent transfers, impersonate executives and run romance and investment scams at scale. The technology that makes a helpful voice assistant also makes a convincing fake of your CFO.

The lesson: authenticity can no longer be assumed from audio or video. The practical defences are dull and effective — call-back verification on payment instructions, code words for sensitive requests, and a default of “verify before you act.” Our sister project counterAI publishes detailed defensive playbooks on exactly this.

Pattern 2 — scaled deception: scams and misinformation

AI lowers the cost of producing convincing text, images and sites to near zero. That has industrialised phishing, fake reviews, fraudulent stores and misinformation. The problem isn’t any single fake — it’s the volume, which overwhelms the old instinct that “it looks professional, so it’s probably real.”

The lesson: polish is no longer a trust signal. For content authenticity specifically, AI detectors help — as one input, not a verdict.

Pattern 3 — taking what isn’t yours: data and copyright

A large share of AI controversy is about inputs: training on copyrighted books, art and code, or scraping personal data without a clear legal basis. This has produced major lawsuits and regulatory scrutiny, and it is a live business risk for AI companies, not just an ethical debate.

The lesson for investors: an AI company’s data provenance is a real liability question. “Where did the training data come from, and do they have the rights?” is one of the due-diligence questions in our AI-washing guide for exactly this reason.

Pattern 4 — surveillance and privacy overreach

AI makes mass analysis of faces, voices and behaviour cheap. Used without consent or proportionality, that tips into surveillance — by employers, platforms or states. Much of the emerging regulation, including the EU’s AI Act, is a direct response to this category.

The lesson: regulatory exposure is now part of the AI business model. Companies built on practices regulators are moving against carry a risk that doesn’t show up in a demo.

Pattern 5 — saying more than you do: AI-washing

The quietest abuse is commercial overstatement: claiming AI capabilities that are thin or absent to win customers or investment. It rarely makes dramatic headlines, but regulators — including securities regulators — have begun treating false AI claims as actionable, not just marketing.

The lesson: this is the abuse most relevant to investors, and the most checkable. It is the entire subject of our verification guide and company verifier.

The thread that ties them together

Across all five patterns, the abuse is never “AI exists.” It is deception, infringement or harm — old wrongs with a powerful new engine. That reframing is useful because it tells you where to put your guard: not against the technology, but against the specific ways people use it to mislead.

For users that means verifying before trusting. For investors it means reading the filings before believing the story. The same discipline, pointed in two directions.

---

Educational content. We describe documented categories of abuse and avoid factual claims of wrongdoing about specific named companies. Not legal or investment advice.