AI & Media
How much money is lost to AI hallucinations every day?
A live estimate of global business losses attributed to AI hallucinations and errors
Roughly $2,130 in estimated global losses every second. (estimated, same model as the counter)
Source: Forrester Research "The Business Case For AI Investment" 2024. View on dashboard →
How do AI hallucinations translate into economic losses?
When AI outputs are wrong or fabricated, the damage is real. Companies using AI for research, legal work, finance, or customer service pay for bad decisions, rework, lawsuits, and reputational hits. Forrester put the global business cost at $67.4 billion in 2024, and it's rising as more firms adopt AI. That's about $2,130 in losses every second worldwide.
What this means for you
Forrester estimated $67 billion in annual business losses attributable to AI hallucinations in 2024, but that number arrives in small increments: a contract clause a lawyer approved without checking the AI's citations, a medical summary a nurse distributed without verifying the dosage, a financial analysis a team built strategy around.
If you use AI tools at work, the cost question is personal. Studies find that knowledge workers spend 30+ minutes per day verifying AI outputs, time that offsets productivity gains. For organisations rolling out AI without verification protocols, the hidden cost is often larger than the efficiency gain.
The practical response is not to stop using AI, but to match verification effort to stakes. Low-stakes tasks (drafting, summarising, brainstorming) can tolerate a hallucination. High-stakes outputs (contracts, medical protocols, financial projections, published journalism) require explicit fact-checking workflows before any AI-assisted content is used.
The $67 billion problem: when AI confidence becomes corporate liability
The business cost of confidence without accuracy
AI hallucinations create economic damage across every sector. In healthcare, incorrect dosage information could harm patients. In law, fabricated case citations result in sanctions and lost cases. In finance, hallucinated analyst data leads to bad investment decisions. In marketing, AI-generated false claims expose companies to regulatory action. The Forrester $67.4 billion estimate is conservative, it captures only documented, attributable losses, not the vast majority that go untracked.
Liability is catching up
The Air Canada and Mata v. Avianca court cases in 2023-2024 established a precedent: companies are liable for what their AI systems say, even when the AI operates autonomously. This shifts the cost of hallucinations from a hidden productivity drain to an explicit legal and financial liability. As case law develops, the economic cost of unverified AI outputs will be more visibly priced into enterprise AI adoption decisions, potentially creating the financial incentive for higher-accuracy, lower-hallucination models to command significant premium pricing.
The financial burden: what studies find
Forrester Research estimated global business cost of AI hallucinations at $67.4 billion in 2024
ForresterA US attorney was fined in 2023 for submitting an AI-generated legal brief with entirely fabricated case citations
Stanford University Human-Centered AIAir Canada lost a legal case in 2024 when its AI chatbot gave false information about bereavement fares
Stanford University Human-Centered AIAI false answers vs. their financial cost, today
Economic cost of AI errors: trend 2023-2024
The business cost of AI hallucinations grew from roughly $40 billion in 2023 to $67.4 billion in 2024, as AI adoption in professional settings expanded faster than verification practices could keep up.
| Year | Rate | Est. per day | Context |
|---|---|---|---|
| 2023 | $1K/sec | $110M | LLMs just entering professional use |
| 2024 | $2K/sec | $184M | AI in professional use grows rapidly; figure from vendor-commissioned study |
| 2026 (forecast) | $3K/sec | $274M | AI adoption accelerates in enterprise |
How AI hallucination cost became measurable
- 2023US attorney Michael Cohen sanctioned for submitting AI-generated fabricated legal citations in federal court
- 2024Air Canada loses court case over AI chatbot providing false bereavement fare policy; must honour hallucinated terms
- 2024Forrester estimates AI hallucination cost at $67.4B globally; growing with AI adoption
Industry research on AI error costs
| Year | Finding | Value | Source |
|---|---|---|---|
| 2023 | Forrester: AI hallucinations cost businesses ~$40B globally in 2023 in direct and indirect losses | 40.0B USD annual global cost | Forrester |
| 2024 | Forrester: Business cost of AI hallucinations reaches $67.4B globally in 2024 | 67.4B USD annual global cost | Forrester |
| 2024 | Stanford HAI AI Index 2025: nearly 90% of notable AI models in 2024 originated from industry (vs 60% in 2023); training compute for notable models doubles approximately every five months; inference costs falling rapidly | 90 % of notable AI models from industry (2024) | Stanford University Human-Centered AI |
The $67 billion figure in perspective
Every two seconds, the business world loses the equivalent of a US worker's monthly paycheck to AI errors.
$67.4 billion a year is nearly three times NASA's entire annual budget.
At $2,130 a second, the daily total exceeds $184 million. That's enough to fully fund a large hospital for an entire year.
How the number is calculated
The $2,130/sec live rate is derived from Forrester Research's 2024 estimate of $67.4 billion in global annual business losses from AI hallucinations ($67.4B ÷ 31,557,600 seconds/year). Important caveat: this Forrester report was commissioned by SnapLogic, an AI integration vendor with a commercial interest in the findings. No independent study has replicated this specific figure. The datapoint is marked as estimated-historical for this reason. The figure covers documented, attributable losses only; Forrester notes the true cost is likely higher.
Sources: Forrester - State of Generative AI 2024. Methodology →
Frequently asked questions
- How much do AI hallucinations cost businesses?
- Forrester Research estimated the global business cost of AI hallucinations at $67.4 billion in 2024. This includes costs from incorrect decisions, rework, legal exposure, and lost customer trust.
- What types of costs does AI hallucination create?
- Key cost categories include: time spent by employees verifying AI outputs and correcting errors, incorrect business decisions made on hallucinated data, legal liability from AI-generated misinformation (e.g. fabricated case citations in court filings), and reputational damage from AI-produced errors published publicly.
How the AI hallucination cost estimate is sourced
The $67.4 billion figure comes from Forrester Research's 2024 GenAI business impact report. Forrester bases cost estimates on enterprise surveys across North America, Europe, and APAC. However, this report was commissioned by SnapLogic, an AI integration vendor, which is a conflict of interest. No independent peer-reviewed study has confirmed a comparable figure. It remains the most-cited public estimate for AI hallucination economic impact, but should be read as a vendor-sponsored upper-bound estimate rather than an independent finding.
Explore related: AI hallucinations per day - Hours lost to AI rework - AI assistant requests, and the live AnythingCounter dashboard.