Is my "as-is" clause
really as-is?

By Jonathan Kleiman, Barrister & Solicitor · Published August 2026

Suppose you sell a device whose central purpose is keeping cryptographic keys secure. A defect in the firmware you shipped makes some of those keys guessable. Your terms of sale say the product is provided "as-is," disclaim every statutory warranty, and cap your liability at the price of the device. Does that end the matter?

Not necessarily. And Coinkite, the Toronto company behind the Coldcard hardware wallet, may soon find out how much protection those clauses actually provide.

What actually happened with the Coldcard firmware?

According to Coinkite's own disclosures and independent analysis by Block's engineering team, a build error introduced in March 2021 caused Coldcard devices to generate wallet seeds using a predictable software randomiser instead of the hardware random number generator — leaving those seeds guessable for more than five years before anyone noticed. The facts below are drawn from named sources, and each is reported rather than asserted.

In its own technical backgrounder, dated July 30 and updated August 1, 2026, Coinkite explained that its 2021 migration to libNgU — an embedded MicroPython library exposing Bitcoin primitives — changed how seed generation drew its randomness. A preprocessor guard tested only whether the macro MICROPY_HW_ENABLE_RNG was defined, not whether its value was non-zero. Coinkite had deliberately set it to zero because it had written its own hardware wrapper. The result, as Block's engineering team independently documented, was that the build silently fell back to MicroPython's non-cryptographic software generator, seeded largely from non-secret device data, instead of the STM32 hardware peripheral.

On the numbers, the two analyses measure somewhat different things. Coinkite reported that affected Mk2 and Mk3 seeds had roughly 40 bits of effective entropy, and estimated an effective search space of about 72 bits for affected Mk4, Mk5 and Q seeds under its assumptions. Block described the newer devices differently, concluding that the secure-element reseed contributes at most 32 securely distinguished bits once the fallback state and call history are fixed, while identifying larger upper bounds where other device and timing variables remain unknown. The figures are therefore not directly comparable — but both analyses conclude that the intended 128-bit security level was not achieved. The distinction that matters is that this was not a break of the underlying cryptography. It was a collapse of the search space to a size a determined attacker could grind through.

Galaxy Research mapped the first wave on-chain: 1,196 addresses drained in full for 1,082.65 BTC, roughly $70.2 million, in a 41-minute window on July 30, 2026. Every sweep paid an identical hardcoded fee of 30.0 sat/vB — a 30-to-75-times overpay against the median that week — and left no change output. Galaxy found no other Bitcoin transactions in the previous 30 days carrying that signature, and read the pattern as consistent with an automated tool spending keys it already held rather than owners moving their own funds. It also observed that the first wave preceded the vendor's public advisory by roughly 30 hours.

The totals moved quickly. CoinDesk, reporting Galaxy's tracking, put losses at 1,367 BTC — about $88.6 million — across 4,585 addresses as of August 2. By August 3, Fortune reported four waves affecting more than 5,200 addresses and roughly 1,816 BTC, worth close to $116 million.

Coinkite's security advisory, published July 30 and updated August 1, identified the affected firmware — Mk2 and Mk3 "versions 4.0.1 through 4.1.9 inclusive," Mk4 and Mk5 before 5.6.0, and Q before 1.5.0Q, with separate thresholds for the Edge builds (6.6.0X and 6.6.0QX) — and stated that TAPSIGNER, OPENDIME and SATSCARD "are not affected by this bug as they are different codebases." The single most important line for owners is this one: "Updating the firmware does not change or repair an existing seed." A patch fixes future seed generation. It does nothing for a weak seed already in the wild, which is why the company's guidance is to move funds to a new seed generated on fixed firmware.

One small discrepancy between the two primary sources, noted for completeness: Coinkite's advisory begins the affected Mk2/Mk3 range at version 4.0.1, while Block's analysis states that "the change first appeared in released firmware v4.0.0 on March 17, 2021" and gives the range as v4.0.0 to v4.1.9. Nothing in the analysis below turns on which is right.

What the published terms of sale actually say

The contractual side is public. Coinkite's Terms of Sale, posted as version 3.3 and last revised November 27, 2024, contain the structure this article is about. Products are "sold on an 'as-is' basis" and the company disclaims "all representations, warranties and conditions." Aggregate liability is "limited to direct damages not to exceed the purchase price we have received for such Products." Those limitations apply "irrespective of the nature of the cause of action," expressly "including... negligence, tort, or any other legal theory." Any proceeding "must be brought within one year after the event which is the subject of the proceeding has occurred." And the terms carry the standard savings language that because "some jurisdictions do not allow" such exclusions, they "may not apply to you."

Two features of that document matter more than they might appear. The terms are "governed by and construed under the laws of the Province of Ontario," and the buyer "irrevocably attorns to the exclusive jurisdiction of the courts of the Province of Ontario." So this is not a case where an Ontario analysis has to be argued for. The parties have chosen it. Everything below proceeds on that footing, with the caveat that I have not seen which version of the terms would govern any particular purchase — a device bought in 2021 was not bought under a document revised in 2024.

That is the fact pattern. Now the law.

Is this a consumer sale or a commercial one? The fork that decides everything

Whether an "as-is" clause survives at all usually turns on a single question that has nothing to do with how the clause is drafted: was the buyer a consumer? In a business-to-business sale, Ontario's Sale of Goods Act lets the parties contract out of the implied warranties. In a consumer sale, the Consumer Protection Act, 2002 deems those same warranties in and declares any term purporting to negate them void. Same clause, same product, same page of the same terms of sale — and two completely different outcomes depending on who clicked "buy."

The commercial side: as-is is on solid ground

If you are selling to another business, the starting point is section 53 of the Sale of Goods Act, and it is about as permissive as statutory language gets: where a right, duty or liability "would arise under a contract of sale by implication of law, it may be negatived or varied by express agreement or by the course of dealing between the parties, or by usage."

That is the provision your as-is clause is standing on. It means the implied condition of merchantable quality and the implied condition of fitness for a particular purpose — both found in section 15 — are defaults, not floors. Say clearly enough that they do not apply, and between two commercial parties, they generally do not.

The current authority on what "express agreement" requires is Earthco Soil Mixtures Inc. v. Pine Valley Enterprises Inc., 2024 SCC 20, and it is worth getting the standard right because the Court of Appeal's stricter version was overturned. The Supreme Court held that an agreement is express where it is "made in distinct and explicit terms and not left to inference" — it "cannot be implied, inferred or imputed from conduct," and "neither silence nor omission will suffice." But that is a requirement about being stated in words, not about using particular ones: "s. 53 does not require express language; there is no requirement for particular magic words." You do not have to name the statutory condition you are excluding.

The Court also emphasised that the "agreement" half of section 53 "is often the crux of the matter," requiring a meeting of the minds about which rights or liabilities are being changed and how — assessed by ordinary contractual interpretation, reading the words in their factual matrix with the goal of finding the parties' objective intention. The practical upshot for a drafter is that you do not need incantations, but you do need the exclusion actually written down and specific enough that a court can see what the parties agreed to change.

So if a hardware wallet is sold to a fund, a corporate treasury, or a business buying custody equipment for its own operations, an as-is clause of the kind quoted above is doing real work.

The consumer side: the warranty disclaimer is void to that extent

Now change one fact. The buyer is an individual, buying one device, to hold their own bitcoin.

Section 9(2) of the Consumer Protection Act, 2002 deems the implied conditions and warranties under the Sale of Goods Act to apply to goods supplied under a consumer agreement. Section 9(3) then does the thing that catches sellers out: "Any term or acknowledgement, whether part of the consumer agreement or not, that purports to negate or vary any implied condition or warranty under the Sale of Goods Act or any deemed condition or warranty under this Act is void."

Read that again with a drafter's eye. Not "unenforceable if unreasonable." Not "subject to review." Void. And section 9(4) closes the escape hatch: the offending term is severable, and it "shall not be evidence of circumstances showing an intent that the deemed or implied warranty or condition does not apply." You cannot even use your own clause as evidence of what the parties meant. Section 7(1) sits behind both: the rights under the Act "apply despite any agreement or waiver to the contrary."

There is no drafting solution to section 9(3). Bolder capitals do not help. A more explicit acknowledgement clause does not help — subsection (3) reaches acknowledgements "whether part of the consumer agreement or not," which is precisely aimed at the seller who tries to have the buyer agree that the warranties never applied. And Earthco is no answer either: it tells you how to draft an effective express agreement under the Sale of Goods Act, but section 9(3) voids the term no matter how expertly it is drafted.

Be precise about what that does and does not accomplish, because this is where commentary usually overshoots. Section 9(3) voids the offending term to the extent it negates or varies the implied conditions, and section 9(4) makes it severable. It does not delete the rest of the contract. Section 9(3) does not automatically invalidate a separate liability cap, risk acknowledgement or procedural term — each has to be considered on its own, including whether, in its particular operation, that term itself purports to negate or vary the statutory protection. (As the next section explains, whether a cap can do that indirectly is an open question.) So the accurate statement is not that a consumer's as-is clause is meaningless. It is that the same warranty disclaimer may operate in a commercial sale and be ineffective in a consumer one — leaving a device sold with a deemed condition of merchantable quality, and a cap potentially standing in the way of a substantial remedy for breaching it.

That distinction is the hinge of the whole analysis. Defeating the disclaimer gets a consumer buyer a warranty; it does not get them damages. The cap is the next fight, and a harder one.

Who counts as a consumer

Section 1 defines a consumer as "an individual acting for personal, family or household purposes," expressly excluding "a person who is acting for business purposes." That is an activity test, not a sophistication test. A software engineer who understands elliptic-curve cryptography considerably better than the seller's lawyers is still a consumer if they bought the device to secure their own savings. Expertise is irrelevant; purpose is everything.

Which is why a product like a personal hardware wallet sits awkwardly for the seller. Its natural market is individuals securing their own holdings — the paradigm consumer purchase — even though the same SKU also goes to businesses under the same terms of sale.

The arbitration clause and the class-action waiver: the same fork again

There is a third place where the same split appears, and it is the one drafters are least likely to have thought about. Coinkite's published terms provide that disputes will, "at our sole option, be settled by final and binding arbitration" in Toronto, and that the buyer agrees "to waive any right you may have to commence or participate in any class action against us." The Ontario attornment clause then applies only to a claim "that is not subject to arbitration."

Between businesses, that is a conventional and generally effective package. Against a consumer, two provisions of the Consumer Protection Act, 2002 meet it head-on. Section 7(2) provides that a term requiring disputes to go to arbitration "is invalid insofar as it prevents a consumer from exercising a right to commence an action in the Superior Court of Justice given under this Act." Section 8(1) provides that a consumer may commence or join a class proceeding "despite any term or acknowledgment" purporting to prevent it.

So the arbitration option and the class waiver are both real and both effective — against your commercial customers. Against your consumer customers they are, to the extent they block statutory rights, unenforceable by direct operation of the statute. That is the same lesson as the warranty disclaimer, arriving through a completely different clause: the document is not badly drafted, it is simply doing two different jobs depending on who is on the other side.

A note on which Act applies

Ontario has passed a new Consumer Protection Act, 2023, which will repeal the 2002 Act. As of writing it has not been proclaimed in force — e-Laws still carries the repeal as taking effect on a day to be named by proclamation. For any sale made to date, including anything sold in 2021 or in 2026, the 2002 Act is the one that governs.

What happens when "as-is" meets a latent defect in the core function?

Merchantable quality and fitness for purpose are assessed as at the time of sale, not when the defect surfaces. If a latent defect meant the goods were not merchantable when they were delivered, the fact that the problem stayed hidden for years does not cure the breach. This is the point that does the most work in a latent-defect case, and it is the one sellers most often misunderstand.

Paragraph 2 of section 15 of the Sale of Goods Act provides that where goods are bought by description from a seller who deals in goods of that description, "there is an implied condition that the goods will be of merchantable quality." There is a carve-out for defects that an examination by the buyer ought to have revealed — but that carve-out is no help against a flaw buried in a compiled firmware build. No buyer examination reveals a preprocessor guard testing the wrong condition.

Apply that to a device sold to keep private keys secret. If the seed generation was weakened from the moment the firmware shipped, the argument is not that the device deteriorated or was misused. It is that the device never had the quality it was sold as having, and that the buyer's five years of apparent security were an illusion the whole time. Whether that argument succeeds would depend on evidence and on findings no court has made here. But it is the shape of the claim.

The trade-name proviso, which sellers will run hard

There is a real counter-argument and it deserves setting out. Paragraph 1 of section 15 creates the implied condition of fitness for a particular purpose where the buyer makes that purpose known and relies on the seller's skill or judgment — but it ends with a proviso: "in the case of a contract for the sale of a specified article under its patent or other trade name there is no implied condition as to its fitness for any particular purpose."

A hardware wallet is bought by trade name. A seller will say the proviso removes the fitness limb outright. The answer, historically, is that courts have read the proviso narrowly: it bites where the buyer insisted on a branded article for its own sake, rather than relying on the seller's skill and judgment to supply something that would do the job. A buyer who chose a product because the manufacturer held it out as securing keys is relying on skill and judgment, not merely ordering by name.

Two things follow for a drafter. The proviso is a live argument and it is worth knowing it exists. But it only touches the fitness limb — the merchantable quality condition in paragraph 2 is untouched by it, and in a case like this that is the stronger limb anyway.

Can a liability cap survive when the loss dwarfs the price?

Often, yes — and this is where commentary about as-is clauses tends to go wrong. Foreseeability and the enforceability of a liability cap are two different questions, and Ontario courts have repeatedly enforced caps that were wildly disproportionate to the loss they excluded. If you are the drafter, this is the most protective part of your terms. If you are the buyer, it is the hardest obstacle.

Start by separating two things that are easy to run together. The rule in Hadley v. Baxendale divides recoverable damages into losses arising naturally from the breach and losses both parties reasonably contemplated when contracting. That rule decides which losses would otherwise be recoverable. It says nothing about whether a valid limitation clause can cap them — and a valid clause can cap even direct, entirely foreseeable losses.

On the first question, a buyer is in reasonable shape. Ask what a hardware wallet is for. It has one job. If the keys it generates turn out to be guessable and the holdings are swept, that is not a remote consequence nobody contemplated — it is the precise risk the product exists to address. A buyer would have a strong argument that the loss was not too remote. Our guide to breach of contract damages in Ontario covers how remoteness works more generally.

Then comes the separate question, and the answer is less comfortable than the intuition suggests.

Ontario courts enforce disproportionate caps

The case a buyer has to deal with is Deswal v. ADT LLC, 2021 ONCA 475. Homeowners had a residential alarm services agreement whose terms said "ADT is not an insurer" and limited liability to 10% of the annual service charge or $250, whichever was greater. Their house was burgled; they filed a proof of loss of $139,330.73 and sued for $500,000. The Court of Appeal upheld the limitation clause. It was not unconscionable — the restriction was clearly disclosed rather than hidden, appeared prominently near the signature line, and came with an acknowledgement. And public policy favoured enforcement, because an alarm provider is not an insurer and its monitoring fee bears no relationship to the risk being insured against.

That reasoning transfers to hardware uncomfortably well. A vendor selling a device for a couple of hundred dollars is not pricing in the value of whatever the buyer chooses to secure with it, has no way to underwrite that risk, and never held itself out as an insurer of it. Deswal applied that reasoning in a residential consumer agreement, following earlier commercial alarm-contract decisions including Fraser Jewellers (1982) Ltd. v. Dominion Electric Protection Co. (1997), 34 O.R. (3d) 1 (C.A.) and Suhaag Jewellers Ltd. v. Alarm Factory Inc., 2016 ONCA 33 — both of which involved jewellery businesses rather than consumers. Together the cases show that Ontario courts may enforce sharply disproportionate caps where the agreement clearly allocates the insured risk away from the provider.

There are real distinctions, and they are worth stating precisely rather than hopefully. Deswal concerned a contract for services, so the mechanism at the centre of this article — section 9(2) of the Consumer Protection Act, 2002 deeming the Sale of Goods Act conditions into a sale of goods — did not arise in the same way. That distinction has Supreme Court backing: in ter Neuzen v. Korn, [1995] 3 S.C.R. 674, the Court held that "in order for the Sale of Goods Act to apply, a contract must be primarily for the purpose of selling goods," and that "if the sale of a good is merely incidental to what is primarily a contract for services, then the statute will not imply a warranty." A hardware wallet is unambiguously a sale of goods; alarm monitoring is not. More significantly, the Consumer Protection Act, 2002 was not raised or considered in Deswal at all. A buyer who established a deemed statutory condition and then argued that a cap operating to leave that condition without any meaningful remedy is itself an attempt to "negate or vary" it would be making an argument the Court of Appeal has not yet addressed. That is a genuine argument. It is not a prediction, and nothing in the reported cases suggests it is an easy one.

The drafting lesson is accordingly not "caps fail when the loss is large." It is closer to the opposite. What the enforced clauses in this line of cases have in common is that they were conspicuous rather than buried, sat near the signature, were paired with an explicit statement that the seller is not an insurer, and were coupled with a coherent explanation of why the price could not carry the risk. Those are drafting and presentation choices, and they are within your control.

Are exclusion clauses trump cards? The Tercon framework

No. In Tercon Contractors Ltd. v. British Columbia, 2010 SCC 4, the Supreme Court laid the doctrine of fundamental breach "to rest" and replaced it with a three-part analysis — does the clause apply on its wording, was it unconscionable when made, and is there an overriding public policy reason to refuse enforcement. One point of accuracy, because it is frequently misdescribed: the framework is set out most fully in Binnie J.'s dissent, but it is unanimous law. As the Court's own headnote records, "The Court agreed on the appropriate framework of analysis but divided on the applicability of the exclusion clause to the facts."

The three questions, in the Court's words:

1. Does the clause apply at all? "The first issue is whether, as a matter of interpretation, the exclusion clause even applies to the circumstances established in evidence." This is a question of the parties' intention as expressed in the contract — and in practice it is where most exclusion-clause cases are actually decided. A clause disclaiming warranties of quality may or may not, on its wording, reach a claim that the product was negligently designed. Ambiguity is construed against the party relying on the clause.
2. Was it unconscionable when the contract was made? Not unfair in hindsight — unconscionable at formation. The Supreme Court's modern statement of that test comes from Uber Technologies Inc. v. Heller, 2020 SCC 16, which requires two elements: an inequality of bargaining power, and a resulting improvident bargain. Notably for anyone selling through click-through terms, the Court in Uber found inequality where the clause "was part of an unnegotiated standard form contract," there was "a significant gulf in sophistication between the parties," and the affected party could not be expected to appreciate the clause's legal and financial implications. A consumer accepting terms of sale at checkout is a good deal closer to that description than two negotiating businesses are.
3. Is there an overriding public policy reason not to enforce it? This branch is deliberately narrow. The burden sits on the party resisting the clause to show "an abuse of the freedom of contract that outweighs the very strong public interest in their enforcement," and the Court offered "conduct approaching serious criminality or egregious fraud" as its examples. A negligent build error, however costly, is a long way from that. This is the branch commentators reach for and courts rarely grant.

For a drafter the practical message is that branches two and three are not where your clause lives or dies. Branch one is. The wording of the clause — whether it actually captures the claim being made against you — is the thing you control, and the thing most often drafted carelessly.

What about negligence, and what if you discover the defect yourself?

Two questions sit just outside this article and are substantial enough to need their own treatment: whether a buyer can get around a contractual cap by pleading negligence, and what a business must actually do once it discovers a defect in a product it has already sold.

The short answers are these. Pleading negligence does not make the contract disappear, and a claim for purely financial loss runs into 1688782 Ontario Inc. v. Maple Leaf Foods Inc., 2020 SCC 35 — no general right to recover pure economic loss, with the negligent-supply-of-shoddy-goods category grounded in the need to avert danger to persons or property. A cap drafted to apply "irrespective of the nature of the cause of action" and naming negligence expressly, as Coinkite's published terms do, is the clear language courts look for. And on the other side of the transaction, a manufacturer's duty to warn is a continuing one — but the Ontario cases establishing it are built on physical danger, which leaves their reach into a purely financial harm genuinely unsettled.

Both are worked through, along with why coordinated disclosure is not a legal safe harbour and how the analysis changes once a flaw is being actively exploited, in the companion guide: do you have to warn customers about a defect you already shipped?

Has the limitation period already expired if the defect is five years old?

Not necessarily. Under section 4 of the Limitations Act, 2002 the basic two-year period runs from the day the claim was discovered, not the day of purchase — so a defect that stayed hidden until July 2026 can produce a claim that is only now starting its clock. This is the section where sellers relax too early and buyers give up too soon.

Section 4 provides that "a proceeding shall not be commenced in respect of a claim after the second anniversary of the day on which the claim was discovered." Section 5(1) then defines discovery as the earlier of the day the claimant first knew of the loss, its cause, the identity of the person responsible and that a proceeding would be an appropriate remedy — and the day "a reasonable person with the abilities and in the circumstances of the person with the claim first ought to have known" those things.

A buyer whose device shipped in 2021 could not reasonably have known that its seed generation was weakened. There was nothing to observe. The flaw sat inside a compiled build, and the wallet behaved exactly as expected until the funds moved. On that footing, discovery arrives when the loss and its cause become knowable, which for most owners means the disclosures at the end of July 2026.

Two qualifications. First, section 5(2) creates a presumption that the claimant knew of the relevant matters "on the day the act or omission on which the claim is based took place, unless the contrary is proved." The burden of displacing that presumption sits with the buyer. In a genuine latent-defect case it is usually dischargeable, but it is a burden, not a formality. Second, section 15 imposes a 15-year ultimate limitation period running from the act or omission itself, regardless of discoverability. Measured from a March 2021 build that outer boundary is not close — but for a business defending a claim about an older product, it is the provision to check first.

The contractual one-year clause, and why it is the fork in miniature

This is where the consumer/commercial distinction produces its cleanest result. Coinkite's published terms require that any proceeding "must be brought within one year after the event which is the subject of the proceeding has occurred." Shortening a limitation period by contract is regulated directly by section 22 of the Limitations Act, 2002.

Section 22(1) provides that a limitation period under the Act "applies despite any agreement to vary or exclude it," subject only to listed exceptions. The exception that permits a period to be varied — and section 22(6) defines "vary" as including "shorten" — applies, in section 22(5), "only in respect of business agreements." Section 22(6) then defines a business agreement as "an agreement made by parties none of whom is a consumer as defined in the Consumer Protection Act, 2002."

So the same clause, once again, does two different things. Against a business buyer, a one-year contractual limitation is something section 22 permits the parties to agree, so a clearly incorporated clause may well be enforceable — subject to the ordinary questions of whether it was incorporated at all and whether it covers the cause of action actually pleaded. Against a consumer buyer, the agreement is not a business agreement, section 22(5) does not apply, and section 22(1) leaves the statutory two-year discoverability period running regardless of what the contract says. A consumer who discovered the problem at the end of July 2026 is not obviously constrained by a one-year clause at all.

For a drafter that is the whole article in a single provision. The clause is not badly drafted. It is simply doing nothing against half your customers, and no amount of redrafting changes that, because the statute keys the exception to who the buyer is rather than to how the term is worded.

Worth noting for completeness: section 15(4)(c) suspends the ultimate period where a defendant "wilfully conceals" the loss or its cause. Whether that provision could ever be engaged in a case like this would depend on evidence about what the seller knew and when — a question this article does not attempt to answer, and one that public reporting cannot settle. It is simply the provision to look at if a buyer alleges that a defect was known and withheld.

You can work through your own dates with our Ontario limitation period calculator, and our guide on how long you have to sue for breach of contract in Ontario covers the discoverability rule in more depth. In a latent-defect case the dates are genuinely contestable, so confirm them with a lawyer rather than relying on a calculation alone.

What this means for your business

If you write as-is clauses, the summary is that your warranty disclaimer does real work against commercial buyers and none against consumers — while your liability cap, the part most people worry about, is probably the sturdiest thing in the document.

If you sell products under as-is terms

• Know which side of the fork each sale sits on. If any meaningful share of your revenue comes from individuals buying for their own use, your disclaimer is void as to those sales and no amount of redrafting changes it. Plan around the exposure instead of papering over it.
• Do not rely on your Ontario address as a shield. Section 2(1) makes your own location a hook into the Act for consumer buyers anywhere.
• Present the cap the way the enforced ones were presented. Deswal, Fraser Jewellers and Suhaag Jewellers upheld caps that were conspicuous rather than buried, sat near the signature, carried an acknowledgement, and said plainly that the seller is not an insurer and its price does not carry that risk. Those are presentation choices you control.
• Draft for branch one. Most exclusion clauses fail because they do not, on their words, capture the claim actually brought. If you intend to exclude liability for negligent design, say that specifically.
• Have a disclosure plan before you need one. Decide in advance who makes the call, how you document it, and what triggers immediate notice rather than an embargo. The trigger that matters most is evidence of active exploitation.

If you bought a product that failed at its core purpose

• An as-is clause is not the end of the analysis — but the cap may be. If you bought as an individual for personal use, section 9(3) may void the warranty disclaimer regardless of how emphatically it is worded. That gets you a warranty. The liability cap is a separate and considerably harder obstacle, and Ontario courts have enforced sharply disproportionate caps in analogous security-service cases.
• The clock probably runs from discovery, not purchase. Do not assume an old product means an expired claim.
• Preserve everything now. Proof of purchase, firmware version, device model, dates, and the transaction record. Evidence decides these cases, and it degrades quickly.
• Get the analysis done properly. Whether any particular claim is worth bringing depends on facts specific to you, and this article does not attempt to answer that.

Key takeaways

• The consumer/commercial fork decides most of it. Sale of Goods Act s. 53 lets commercial parties contract out of the implied warranties, on the Earthco standard; Consumer Protection Act, 2002 s. 9(3) makes any term negating them void in a consumer sale — and s. 2(1) reaches you through your own Ontario address.
• Voiding the disclaimer is not the same as winning. Section 9(4) makes the offending term severable, and s. 9(3) does not automatically invalidate a separate cap or acknowledgement — each is assessed on its own, including whether it negates the statutory protection in its actual operation.
• The fork shows up again in your dispute-resolution clauses. An arbitration option and a class-action waiver work between businesses; CPA ss. 7(2) and 8(1) largely disarm both against consumers.
• Disproportionate caps get enforced in Ontario. Deswal v. ADT, 2021 ONCA 475 upheld a 10%-of-annual-fee cap against a $139,000 burglary loss. Foreseeability under Hadley decides what is recoverable, not whether a cap can limit it.
• Negligence is not the easy way around a cap. Maple Leaf Foods confirms there is no general right to recover pure economic loss, and a cap naming negligence expressly is the clear language courts look for. The duty-to-warn cases are built on danger, not financial harm.
• Discoverability, not purchase, starts the clock. Limitations Act, 2002 s. 4 runs two years from discovery, and s. 22 lets you shorten it by contract only in a business agreement.

Frequently asked questions

Is an "as-is" clause enforceable in Ontario?

It depends almost entirely on who bought the product. In a business-to-business sale, section 53 of the Sale of Goods Act lets parties negative or vary the implied conditions of merchantable quality and fitness for purpose by express agreement. In Earthco Soil Mixtures Inc. v. Pine Valley Enterprises Inc., 2024 SCC 20, the Supreme Court held that such an agreement must be made in distinct and explicit terms rather than left to inference — but that section 53 "does not require express language" and there is "no requirement for particular magic words," so you need not name the statutory condition you are excluding. In a consumer sale, section 9(3) of the Consumer Protection Act, 2002 makes void any term that purports to negate or vary those implied conditions — no reasonableness test. So the same warranty disclaimer may operate in a commercial sale and be ineffective in a consumer one. Note what that does not do: section 9(3) voids the term to the extent it negates the implied conditions and section 9(4) makes it severable, but it does not automatically invalidate a separate liability cap or the rest of the contract. Those provisions must be considered independently, including whether the cap, in its actual operation, itself purports to negate or vary the statutory protection.

Can I disclaim the Sale of Goods Act warranties when selling to consumers in Ontario?

No. Section 9(2) of the Consumer Protection Act, 2002 deems the Sale of Goods Act implied conditions and warranties to apply to goods supplied under a consumer agreement, and section 9(3) makes void "any term or acknowledgement, whether part of the consumer agreement or not, that purports to negate or vary" them. Section 9(4) goes further: the offending term is severable and cannot even be used as evidence that the parties intended the warranty not to apply. Section 7(1) provides that rights under the Act apply despite any agreement or waiver to the contrary. There is no drafting workaround — capital letters, a separate acknowledgement, or a signature line beside the disclaimer all fail for the same reason.

Does the Ontario Consumer Protection Act apply if my customers are outside Ontario?

It may well, and this is the point Ontario sellers most often miss. Section 2(1) says the Act applies to consumer transactions "if the consumer or the person engaging in the transaction with the consumer is located in Ontario when the transaction takes place." Your own Ontario location is enough to bring the transaction within the Act — the buyer does not have to be here. An Ontario company selling online to consumers anywhere should assume it is inside the Act, and should not assume a governing-law clause selecting a friendlier jurisdiction solves the problem, given section 7(1). This is worth specific advice if you sell cross-border at volume.

Can a liability cap limit my exposure to the price of the product?

Frequently, yes — more often than people expect. Two separate questions get run together here. Hadley v. Baxendale decides which losses would otherwise be recoverable, and where a product exists for one purpose the loss from a failure of that purpose is generally not too remote. But that says nothing about whether a valid limitation clause can cap those losses, and a valid clause can cap even direct, foreseeable ones. In Deswal v. ADT LLC, 2021 ONCA 475, the Ontario Court of Appeal enforced a clause limiting liability to 10% of the annual service charge or $250 against a residential burglary loss exceeding $139,000, applying reasoning from earlier commercial alarm-contract cases including Fraser Jewellers and Suhaag Jewellers. The clauses that survive tend to be conspicuous rather than buried, positioned near the signature, and paired with an explicit statement that the seller is not an insurer and its price does not carry that risk.

Can I require arbitration and waive class actions in my terms of sale?

Between businesses, generally yes. Against consumers, Ontario legislates against both. Section 7(2) of the Consumer Protection Act, 2002 provides that a term requiring disputes to go to arbitration is invalid insofar as it prevents a consumer from exercising a right to commence an action in the Superior Court of Justice given under the Act. Section 8(1) provides that a consumer may commence or become a member of a class proceeding despite any term or acknowledgment purporting to prevent it. Section 7(3) does allow the parties to agree to another dispute-resolution process after a dispute has actually arisen — the restriction bites on advance waivers, not on genuine post-dispute settlement choices. So an arbitration option and a class-action waiver can be fully effective against your commercial customers and largely inoperative against your consumer ones.

Does an exclusion clause always fail if the breach is serious enough?

No — Canadian law abandoned that idea. In Tercon Contractors Ltd. v. British Columbia, 2010 SCC 4, the Supreme Court held that the doctrine of fundamental breach should be "laid to rest" and replaced it with a three-part analysis: first, does the clause apply on its wording to what actually happened; second, was it unconscionable at the time the contract was made; and third, is there an overriding public policy reason to refuse enforcement. The Court unanimously agreed on that framework even though it split on the outcome. The third branch is deliberately narrow — the Court gave "conduct approaching serious criminality or egregious fraud" as examples — so an ordinary negligent defect is unlikely to get there. Most real fights are won or lost on the first branch, on the words of the clause itself.

How long does a buyer have to sue if the defect was hidden for years?

Under section 4 of the Limitations Act, 2002, the basic limitation period is two years from the day the claim was discovered — not from the date of purchase. Section 5(1) defines discovery by reference to when the claimant first knew, or a reasonable person in their circumstances first ought to have known, of the loss and its cause. So a defect that stayed hidden for five years does not necessarily produce a claim that expired four years ago. Section 5(2) does presume the claim was discovered on the day the act or omission took place unless the contrary is proved, so the buyer carries the burden of rebutting that presumption. Section 15 adds a 15-year ultimate limitation period measured from the act or omission, which is a hard outer boundary in most cases. Watch for a contractual clause shortening the period: section 22(1) says the statutory period applies despite any agreement to vary it, and section 22(5) permits variation only in a "business agreement" — defined in section 22(6) as an agreement where none of the parties is a consumer. A one-year suit clause can therefore bind a business buyer and do nothing against a consumer. You can work through the timing on our Ontario limitation period calculator, but the dates in a latent-defect case are worth confirming with a lawyer.

Final thoughts

The reason this incident is worth a business lawyer's attention has nothing to do with cryptocurrency. It is that it puts a very ordinary contractual structure — as-is, warranties disclaimed, liability capped at the price — under a kind of pressure that ordinary structure rarely faces. Most products fail in small, survivable ways, and an as-is clause absorbs the complaint. Occasionally a product fails completely at the one thing it existed to do, and the clause has to carry weight it was never really built for.

Whether Coinkite bears any legal responsibility for any particular loss is not something this article answers or attempts to. That would depend on the terms applicable to the particular purchase, evidence that is not public, and findings no court has made. What the episode does illustrate, with unusual clarity, is how much of the protection businesses believe they have bought with three sentences of boilerplate depends on facts they may not control: who the buyer was, where the seller sits, what the product was sold to do, and when the problem became knowable — and how much of it turns out to be sturdier than they feared, in the one clause they worry about most.

There is also a forward-looking point that sellers of software-bearing products should sit with. Coinkite's own framing was that attackers and defenders now have the same AI tools and that on this occasion the tools helped the attackers first. If latent defects in shipped code are going to surface faster and in older products than they used to, then the gap between "we shipped it" and "someone found the flaw" is going to close. The contractual and disclosure planning that follows from that is worth doing before you need it, not after.

If you want your terms of sale pressure-tested — particularly the warranty disclaimer, the liability cap, and whether they work across both consumer and commercial sales — a contract lawyer in Toronto can tell you where the real exposure sits. Call 416-554-1639 or book a consultation.

Related services and guides

Contract drafting, liability and dispute resources from around the site.

Related services

• Contract lawyer in Toronto — drafting and reviewing terms of sale, disclaimers and liability caps
• Business lawyer in Toronto — governance, risk and commercial advice for owner-managed businesses
• Contract dispute lawyer — when the other side says your clause does not protect you
• Commercial litigation lawyer — disputes that have moved past the negotiating stage

Related guides

• Do you have to warn customers about a defect you shipped? — the companion guide on the duty to warn, negligence and disclosure timing
• Breach of contract damages in Ontario — how losses are measured and what is too remote
• What makes a contract legally binding — the elements behind every terms-of-sale document
• How long you have to sue for breach of contract — the two-year clock and the discoverability rule
• How to enforce a contract in Ontario — the practical mechanics once a dispute is real
• When your AI assistant signs a bad deal — the other side of AI risk in commercial contracting
• Ontario limitation period calculator — work out your own dates under the Limitations Act, 2002