Bitcoin's best-documented problems concern how people use it: Bank of Canada surveys found that about half of Canadian Bitcoin adopters had run into price crashes, lost access to funds, scams, or data breaches, and the US Energy Information Administration (EIA) estimated that crypto mining used between 0.6% and 2.3% of US electricity in 2023. Researchers have also shown that machine learning can pick out some illicit Bitcoin transactions, with real limits. The material below is informational and gives no investment advice.
Key takeaways
- In Bank of Canada surveys since 2018, about 90% of Canadians had heard of Bitcoin and only 5% owned it.
- About half of past and current Canadian adopters reported price crashes, lost access to funds, scams, or data breaches.
- The EIA's preliminary estimate put US crypto mining at 25 to 91 TWh in 2023, or 0.6% to 2.3% of US electricity use.
- A 2019 research paper tested machine learning on labeled Bitcoin transactions; the best model flagged illicit activity with high precision but missed about a third of cases.
- Nothing here predicts prices or recommends any action with Bitcoin.
What Canadian survey data say about ownership and harm
Bank of Canada Staff Discussion Paper 2022-10 analyzed Bitcoin and cash-alternative survey data collected annually from 2016 to 2020. In the surveys since 2018, about 90% of respondents had heard of Bitcoin, while only 5% owned it. Investment remained the most common reason for owning it, and most owners had obtained it through web or mobile exchanges.
The paper also reports that adopters, past and current, were exposed to financial and security risks: about half experienced price crashes, loss of access to funds, scams, or data breaches. That figure bundles four different problems into one, so it cannot say how many people suffered permanent losses. The data end in 2020, and no later survey from the same series was found for this page, so the numbers describe that period and not October 2026.
Electricity use and why the estimates vary
Bitcoin mining runs on specialized computers, and its electricity demand has drawn regulatory attention. In an article published on February 1, 2024, the EIA applied the Cambridge Bitcoin Electricity Consumption Index and an assumed US share of about 38% of global mining. That produced a preliminary range of 25 to 91 TWh for 2023, set against US demand of 3,900 TWh. A separate bottom-up count of 137 facilities, with usable power data for 101 of them, totaled about 10,275 MW of maximum draw, or about 70 TWh a year at an assumed 80% utilization.
The EIA called all of this preliminary. Facility locations and power draw are incompletely known, and miners relocate toward cheaper power. The agency has since discontinued its emergency survey of mining operations, so the 2023 estimate has not been replaced by a measured national figure. Any single energy number for Bitcoin should be read as a modeled range.
Irreversible payments and common scam patterns
Because confirmed Bitcoin transfers are generally not reversible, mistakes and fraud are costly. The US Federal Trade Commission states that cryptocurrency payments typically are not reversible and lack the legal protections that cards offer. The CAFC, Canada's national anti-fraud agency, lists red flags that include fake crypto websites, unsolicited offers, and demands for crypto payment, and it adds that official agencies never ask for payment in crypto. A companion page on this site covers scam patterns and an Ontario Securities Commission experiment on AI-enhanced scams in detail.
Can machine learning trace illicit Bitcoin activity?
Every Bitcoin transaction is recorded publicly under pseudonymous addresses, which gives investigators data to analyze. In a 2019 paper, Weber and colleagues released the Elliptic Data Set: 203,769 Bitcoin transactions across 49 time steps, of which 4,545 were labeled illicit, 42,019 licit, and the rest unlabeled. They compared logistic regression, random forests, multilayer perceptrons, and graph convolutional networks.
A random forest using all 166 features did best on the illicit class, with precision of 0.956 and recall of 0.670. In plain terms, nearly all flagged transactions were truly illicit, but about a third of the illicit ones went unflagged. The authors also discuss a policy tension: anti-money laundering rules carry costs and can exclude people from financial services. The labels came from a single research effort, and illicit patterns shift over time, so results from this dataset do not translate directly into a working compliance system.
Concerns, evidence, and limits
The table below sketches one way to separate what has been measured from what remains open.
| Concern | Documented evidence | Limit of the evidence |
|---|---|---|
| Harm to adopters | About half of Canadian adopters reported crashes, lost access, scams, or breaches (Bank of Canada, data to 2020) | Four problems merged into one figure; data are six years old |
| Electricity use | 25 to 91 TWh for US mining in 2023 (EIA) | Modeled range; survey discontinued |
| Irreversible payments | FTC and CAFC guidance on non-reversible transfers | Describes risk, not frequency |
| Illicit use | Labeled dataset with 4,545 illicit transactions | A small share of one dataset, not a measure of all Bitcoin activity |
| Detection by AI | Random forest recall of 0.670 on the illicit class | Single dataset; labels incomplete |
Information, not advice
This page reports published findings about Bitcoin's risks and about research on detection. It does not judge whether holding or using Bitcoin suits anyone, and it says nothing about future prices. Crypto assets can lose much of their value, and the Bank of Canada data show that problems were common among past adopters. Readers facing a money decision can speak with a registered adviser, check any firm with their provincial securities regulator, and read official guidance from the CAFC or the FTC.
FAQ
How many Canadians owned Bitcoin?
In the Bank of Canada surveys since 2018, about 5% of respondents owned it, while about 90% had heard of it. The series ends in 2020.
How much electricity does Bitcoin use?
No measured national figure exists for the United States. The EIA's preliminary modeled range for 2023 was 25 to 91 TWh, or 0.6% to 2.3% of US electricity consumption.
Can AI identify illegal Bitcoin transactions?
In research settings it can flag some. The 2019 Elliptic study found high precision but incomplete recall, which means missed cases and a need for human review.
Are Bitcoin transactions anonymous?
They are pseudonymous. Addresses appear on a public ledger without names, and that same public record is what researchers use for transaction analysis.
Related reading
- Crypto scams and the AI studies behind them
- How AI and cryptocurrency differ
- Open problems in artificial intelligence
- AI tools used around crypto markets
More in AI in finance.
Last verified: October 2026
Sources
- Bank of Canada, "Bitcoin Awareness, Ownership and Use: 2016-20," Staff Discussion Paper 2022-10, 2022
- US Energy Information Administration, "Tracking electricity consumption from U.S. cryptocurrency mining operations," February 1, 2024: source page
- Weber et al., "Anti-Money Laundering in Bitcoin: Experimenting with Graph Convolutional Networks for Financial Forensics," arXiv:1908.02591, 2019: source page
- US Federal Trade Commission, "What To Know About Cryptocurrency and Scams," consumer advice
- CAFC (Canada's national anti-fraud agency), "Investment" fraud page