The risks of AI — hallucinations, data privacy, and what should never go into a chatbot. 

Detected Manipulations & Suggested Changes

Sensationalism / Appeal to emotion

Using emotionally charged or dramatic wording that can make risks feel larger or more alarming than the underlying evidence strictly supports.

1) "A chatbot is never more dangerous than when it is specific, fluent, and wrong." 2) "Most AI data leaks are not sabotage; they are a helpful employee with a deadline. Ban the tools outright and that employee keeps using them anyway, on a personal phone, on a personal account, with none of the protections above. The stronger move is a short written policy..."

3 suggestions

Replace "A chatbot is never more dangerous than when it is specific, fluent, and wrong" with a more measured formulation such as: "A chatbot can be especially risky when it is specific, fluent, and wrong, because errors are harder to spot."

Clarify the claim about bans by adding nuance and evidence, for example: "In many workplaces, banning the tools outright may push some employees to use them anyway on personal devices, where there are fewer protections. A short written policy can reduce this risk by clarifying approved tools and safe practices."

Avoid absolute terms like "never more dangerous" and instead use probabilistic or conditional language (e.g., "often", "can be", "in many cases").

Oversimplification

Presenting a complex behavioral or policy outcome as if it follows straightforwardly from a single cause, without acknowledging variability or exceptions.

The passage: "Most AI data leaks are not sabotage; they are a helpful employee with a deadline. Ban the tools outright and that employee keeps using them anyway, on a personal phone, on a personal account, with none of the protections above. The stronger move is a short written policy..." suggests that bans will generally drive covert, unsafe use, and that a short policy is categorically the "stronger move" without discussing contexts where strict bans plus monitoring might be effective.

3 suggestions

Qualify the causal claim, for example: "In some organisations, banning the tools outright may lead employees to use them anyway on personal devices, with fewer protections."

Acknowledge alternative approaches and conditions, such as: "Depending on the organisation’s risk tolerance and regulatory environment, a combination of clear policy, training, and technical controls (which may include partial or full bans) can be effective."

Replace the categorical phrase "The stronger move is a short written policy" with a more balanced one like: "For many businesses, a short written policy is more practical and effective than an outright ban."

Framing effect / Mild confirmation bias

Presenting information in a way that consistently supports the author’s preferred approach (adopt AI with safeguards) while giving less space to alternative positions, which can subtly steer readers toward one conclusion.

Throughout the article, the framing assumes that AI "belongs in the Jamaican finance office" and that the key question is how to use it safely, not whether to use it at all. For example: "For nine weeks this column has argued that AI belongs in the Jamaican finance office..." and "Neither risk is a reason to sit the technology out. Both are reasons to use it with rules...". The possibility that some organisations might reasonably choose to avoid or severely limit AI use is mentioned only implicitly and not explored.

3 suggestions

Explicitly acknowledge that some organisations may decide not to adopt AI tools at all, for example: "Some businesses may still decide that, given their risk profile or regulatory obligations, avoiding these tools is the right choice."

Add a brief discussion of scenarios where a more restrictive or no‑AI policy could be justified (e.g., highly regulated sectors, extremely sensitive data).

Rephrase "Neither risk is a reason to sit the technology out" to something like: "For many businesses, these risks are not a reason to avoid the technology entirely, but they are strong reasons to use it under clear rules."