iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
AI coverage became much more prominent and more focused on danger after ChatGPT launched on November 30, 2022. That shift is measurable in a study of selected English-language news sites—but it does not prove that the risks being reported are imaginary, or that every warning is justified. To judge whether a claim is alarmist, separate what has happened from what experts think could happen, and look at how specific and strong the evidence is.
What changed in AI news coverage after ChatGPT?
Deißler and colleagues analyzed more than 49,000 sentences from 5,846 articles published by 18 highly visited English-language news sites. Their comparison covered the six months before and the six months after ChatGPT’s November 30, 2022 launch. They report that media attention to AI rose tenfold in the six months after launch, from an already high level.
The study also found a shift in how stories framed AI. In its sentence-level analysis, the share of danger-related material that identified AI as a source of danger rose from 35% to 64% for annotator A1, and from 31% to 67.5% for annotator A2. The share of articles containing at least one of the study’s selected frames increased from 37.7% to 44.8%.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →These results describe changes in this sample’s coverage, not a global census of news, a measure of what readers believe, or a test of whether each risk claim is true. The authors also found more discussion of specific harms, such as fraud and abuse, alongside more broad or unspecified threats. Those are different kinds of reporting: a described misuse is not evidence that a much wider future scenario is already occurring.
#1 Best Overall
Does more alarming coverage mean AI risks are exaggerated?
No. “More alarm-oriented coverage” is a conclusion about framing: news stories more often presented AI as a source of danger. It is not a verdict on whether the danger exists. A real harm can receive disproportionate attention; a plausible risk can be reported cautiously; and a dramatic prediction can be weakly supported even if it concerns a genuine area of uncertainty.
A useful way to assess a claim is to ask four questions: Is it about an observed event or a future possibility? Is it supported by measured evidence, expert judgment, or a theoretical argument? Does it name a mechanism, or describe a broad threat? And does the evidence concern one incident or a wider pattern? Keeping those dimensions separate avoids treating every warning as proof—or every uncertainty as reassurance.
| Claim type | What to look for | What it can establish |
|---|---|---|
| Observed current harm | A documented incident or measured pattern, with a specified mechanism and scope. | That a particular harm occurred or was measured; it does not automatically establish its overall prevalence. |
| Expert judgment | Who was surveyed, how respondents were recruited, what question was asked, and when. | What those respondents believe or expect, not necessarily what will happen. |
| Future scenario | The proposed pathway from today’s capabilities to the predicted outcome, and the assumptions it depends on. | That a scenario is considered possible or worth examining; it is not evidence that the outcome is underway or inevitable. |
What do experts think about catastrophic AI risks?
There is disagreement, but concern is not confined to a fringe in the available survey. Severin Field’s 2025 preprint reports results from 111 AI experts, including safety researchers, academics, and industry professionals. It describes two broad perspectives: AI as a controllable tool and AI as an uncontrollable agent. In that sample, 78% agreed or strongly agreed that technical AI researchers should be concerned about catastrophic risks.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
That percentage is a result from this particular survey, not a representative estimate of all AI experts. Field emailed 1,090 potential participants and reports a 10% response rate; respondents were recruited through AI and machine-learning professional networks. The preprint also reports that 21% of respondents had heard of “instrumental convergence,” a technical idea relevant to some arguments about advanced AI. Survey responses can inform a debate, but the sample, wording, recruitment, and respondents’ definitions all shape what the numbers mean.
Rank #3
How strong is the evidence about AI disinformation?
Generative systems can produce realistic text, voices, images, and video, which can support deception at scale. But the ability to create convincing material is not the same as proof that it has caused large-scale real-world effects. In their 2025 review, journalism and mass-communication researchers Christian von Sikorski and Michael Hameleers describe empirical evidence about AI-driven disinformation’s actual effects as mixed.
Their discussion of Romania’s 2024 presidential election says direct evidence of AI deployment was inconclusive. They argue that it is too early to make a definitive judgment about whether concerns about AI and disinformation are warranted or exaggerated. That qualification matters: a plausible method of manipulation deserves scrutiny, but the method’s existence does not establish how often it is used or how much influence it has.
Expert perceptions also indicate concern, while measuring something different from observed harm. The International Panel on the Information Environment’s 2025 report, describing its 2024 expert survey, says 63% of respondents believed AI-generated videos, voices, images, and text had negatively affected the global information environment; 53% expected a negative impact over the following five years. Those are surveyed experts’ assessments and expectations, not direct measurements of the scale of past or future effects.
Recommended Free Tools
How to judge an alarming AI headline
- Check the time horizon. Does the story describe a documented event, a current capability, or a prediction about a future system?
- Look for the mechanism. A concrete account of fraud or impersonation is easier to assess than a sweeping claim that AI will cause unspecified harm.
- Match the conclusion to the evidence. A survey measures respondents’ views; a case study describes a particular case; neither alone establishes a global trend.
- Notice the scope. Ask whether evidence comes from one incident, a selected set of news sites, or a broader and representative dataset.
- Separate possibility from probability. A credible pathway to harm may justify precautions without showing that the harm is common, imminent, or inevitable.
The evidence supports a clear but limited conclusion: AI received more attention and more danger-focused framing in a large sample of English-language news after ChatGPT’s launch. At the same time, current misuse, disputed real-world effects, and catastrophic future scenarios require different evidence. Neither blanket dismissal nor certainty about worst-case outcomes follows from the studies and reviews discussed here.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
