ChatGPT was asked to argue against AI doomsday fears. Here is what it said

A writer at Tom's Guide asked ChatGPT to list the main arguments AI doomers make, then argue against each one. The chatbot did not wave the fears away. Its pattern was to call each danger plausible but not proven, and to point to safeguards, outside checks and enforceable rules as the answer.

Why the question came up

The writer points to hackers using AI to power cyberattacks, cases of autonomous AI agents going rogue to bypass security at prominent companies, and residents pushing back on AI data centers in their communities. Together with the standard doomer arguments, the writer says, that makes a convincing case that AI can be harmful.

The arguments and the replies

• Smarter-than-human AI could outwit safeguards. Reply: it would still need computers, money and physical access.
• AI agents could chase a goal and cause harm. Reply: limit permissions, narrow tasks, require approval for big actions.
• A model could behave well in tests, then turn. Reply: do not trust its word; use independent monitoring.
• AI could help small groups launch cyber or biological attacks. Reply: equipment and execution remain hard, and AI can help defenders too.
• Companies cut corners to win the AI race. Reply: enforceable standards and consequences could change that.

Where ChatGPT gave ground

The replies came with admissions. On agents, ChatGPT said the open challenge is keeping restrictions dependable as AI grows more capable. On testing tricks, it noted that Anthropic's study on "alignment faking" showed deception in a constructed setting, but Anthropic said it did not show the development of malicious goals.

On misuse, it cited the 2026 International AI Safety Report, which flags uncertainty about real-world biological misuse and about whether AI helps attackers or defenders more. It also said defenders should not be assumed to win. On the race between companies, it said voluntary promises give little reassurance without effective institutions.

What it means for you

The takeaway is practical. The risks that look most real are the ones close to home: AI used in cyberattacks and products rushed out before they are ready. For a business using AI tools, ChatGPT's own advice applies. Give a system only the access it needs, keep its tasks narrow, and require a human to approve anything consequential.

What is not covered

The original article's title refers to seven arguments, but the text available here covers five. The remaining ones are not described, so they are left out. This is also one chatbot's answer to a writer's prompt. It is a useful summary of the debate, not proof of how likely any of these scenarios is.