Akamai v. Limelight: Joint Infringement Requires an Agency Relationship or a Contractual Obligation

By Jason Rantanen and Dennis Crouch

Akamai Technologies, Inc. v. Limelight Networks, Inc. (Fed. Cir. 2010)
Panel: Rader, Linn (author), Prost

Three years ago, in BMC Resources, Inc. v. Paymentech, L.P., 498 F.3d 1373 (Fed. Cir. 2007), the Federal Circuit sharply limited the ability of patent holders to assert claims of joint infringement.  The BMC decision was further bolstered by Muniauction, Inc. v. Thomson Corp. and I-Deal (Fed. Cir. 2008). Akamai v. Limelight expands on the "control or direct" rule announced in BMC and Muniauction, explaining that joint infringement requires an agency relationship or a contractual obligation to carry out the relevant steps. 

Akamai's Claim of Joint Infringement
In order to win a patent infringement case, a patentee must prove that a single entity practiced every element (either literally or by equivalents) of at least one valid, enforceable claim. Joint infringement issues can arise when a patent holder asserts method claims over activities whose steps are not carried out by a single party.  For example, if a patent claims a method involving performing steps A, B, and C, the patent holder may need to argue joint infringement (as opposed to straightforward direct infringement) if steps A and B are performed by one party and step C is performed by a second party. 

In Akamai, while the accused party (Limelight) performed the majority of the steps of the asserted claims, at least one of the steps of each claim was performed by its customers.  The patent, No. 6,108,703, is directed to an improved method for storing web page content.  Conventionally, the entirety of a web page, including both the page itself and embedded content (such as graphics) is stored on a single server, or mirrored in its entirety across multiple servers.  The patents-in-suit claim a new approach, which involves storing only the embedded objects on mirrored servers (called a "Content Delivery Network," or "CDN"), while having the webpage itself continue to reside on the content provider's servers.  In order to make this system work, the claims further require that the object URLs be "tagged" to resolve to the CDN. 

Daniel Lewin and his adviser Thomas Leighton devised these methods (or algorithms) while at MIT where Leighton was the head of MIT's Computer Science and Artificial Intelligence Laboratory.  The pair founded Akamai, and their invention continues to serve as a basis for the company's core business.  Lewin was killed aboard American Airlines flight 11 when it crashed into the north tower of the World Trade Center on September 11, 2001.

Limelight, Akamai's direct competitor in the CDN business, performed most of the claimed steps but had its customers (companies such as Netflix) tag the URLs.  Because some of the steps were performed by Limelight and some by its customers, Akamai relied on a theory of joint liability, arguing that Limelight controls or directs the activities of its customers.  At trial, the jury sided with Akamai and awarded over $40 million in lost-profit-damages. The district court rejected the jury verdict and supplemented its own non-infringement judgment – holding that there was "no material difference between Limelight's interaction with his customers and that of Thomson in Muniauction."

On appeal, Akamai contended that substantial evidence supported the infringement finding, arguing that Limelight "(1) creates and assigns a unique hostname for the content provider; (2) provides explicit step-by-step instructions to perform the tagging and serving claim steps; (3) offers technical assistance to help content providers with their performance of the claim step; and (4) contractually requires content providers to perform the tagging and serving claim steps if they utilize the Limelight service."  Slip Op at 11. 

Important holdings
In affirming the district court's judgment of noninfringement, the CAFC expanded on its ruling in BMC that "joint liability may be found when one party 'control[s] or direct[s]' the activities of another party."  Slip Op. at 9.  After noting the foundational nature of this standard, the Akamai panel further held that "what is essential is not merely the exercise of control or the providing of instructions, but whether the relationship between the parties is such that acts of one may be attributed to the other."   Id. at 12.  The court interpreted this as requiring either an agency relationship or a contractual obligation: "as a matter of Federal Circuit law that there can only be joint infringement when there is an agency relationship between the parties who perform the method steps or when one party is contractually obligated to the other to perform the steps."  Slip Op. at 14.

Analyzing the issue as a question of agency law, the CAFC concluded that no substantial evidence supported a finding that Limelight's customers perform any of the steps of the claimed method as agents for Limelight.  (Citing the Restatement (3rd) of Agency §1.01). The court also rejected Akamai's theory that Limelight's customers are contractually required to perform the tagging step.  Rather, the court stated, the contract does not obligate the customers to perform any of the method steps; instead, it "merely explains that the customer will have to perform the steps if it decides to take advantage of Limelight's service."  Slip Op. at 16.

Note: The court repeated the warning it set forth in BMC Resources, 498 F.3d at 1381, that concerns about the difficulty of proving infringement of claims that must be infringed by multiple parties "by proper claim drafting.  A patentee can usually structure a claim to capture infringement by a single party."  Slip. Op. at 17 (quoting BMC Resources).  Failing this, the court suggested (citing Mark Lemley's Divided Infringement article) that patentees may be able to correct a claim by seeking a reissue patent.

Claim Construction
Akamai also challenged the district court's construction of two claim terms in related patents, contending that both constructions imported limitations from the specification into the claims.  In a fact-intensive analysis, the CAFC rejected Akamai's arguments, concluding that the district court correctly interpreted the disputed terms given the treatment of the invention in the specification and the lack of any contrary evidence in the prosecution history. 

23 thoughts on “Akamai v. Limelight: Joint Infringement Requires an Agency Relationship or a Contractual Obligation

  1. 23

    Can yone square this decision with Sirf?

    Sirf held that for a very similar type of cliams the Court did not even need to reach the issue of joint infringement because the software, not the users, did all the claim steps. The Court basically said that to evaluate who is performing the claimed steps, you assume everything not explicitly recited in the claims is already being done (ready and enabled). So here, one could assume that, for example, the user at clicked and highlighted or whatever and the software did the “tagging” automatically in response to the unclaimed acts of the users.

  2. 22


    Rephrasing from active to passive is not what was bein discussed. Your “after amendment” claim is not different from teh before amendment claim, as you still have first and second actors, evenif ya nested a ‘comprsising” in order to get rid of your “B)”.

    In other words, ya aint fooliin anyone, and ya drop inta the “No clue” bucket.

    While district cases can be seriously flawed (e.g. Sweet), Iza bet that ya are simply misquotiing case law (again). If ya wanna throw an actual cite out there I just mught give ya my oh-so-valuable observations abouts where ya getting off track.

    If ya ask nicely.

  3. 21

    Another dude, I respectfully disagree. I think the issue is whether subtantially the same acts have to be performed by the same actors before and after in order to prove infringement. For this reason, I think it is possible to rephrase certain a claim steps from active to passive without broadening.

    For example, consider two steps:
    A method of processing act X, comprising:
    A) first actor performing act X.
    B) second actor acting on the results from step A.

    After amendment:
    A method of processing act X, wherein act X comprises a first actor having performed act X; the method comprising:
    A) second actor acting on the results.

    All the steps and acts remain, so no broadening. But it is now arguable the claim is infringed by A because only that step is after the phrase “the method comprising.”

    There is a recent district court case to this effect from New York. There the court held background steps set forth in a preample by differenct actors would not be counted for divided infringement purposes, although those steps would still have to be proved to prove infringement.

  4. 19

    The decision is good for the patent system because it reinforces a straightforward statutory line between direct and indirect infringement and promotes the drafting of clearer and better claims.

    And to answer an earlier question, yes, modifying a claim in reissue to fix a divided infringement problem would clearly be “broadening.” Suppose Claim X could not be directly infringed by Company A because one step of that claim is actually performed by another party. Claim X is later amended in reissue or reexamination so that that step is now performed by Company X. For example, a step of a user sending data to a web server is amended to instead recite the web server “receiving” that same data from the user. How is that not broadening? Company X’s activities would not directly infringe before (absent proof of agency or contractual obligation over the user), but now they do? No clue why anyone would say that this isn’t a broadening.

  5. 17

    So if it’s not about the Facts, and I can’t speak then what? I want Justice. I have letters to KT showing that I wanted it to hang over the side of the boat. I never sent drawings to a McCombs. I sent them to KT and then three to JRW when he claimed KT was gone. And that was after the Search. You know all this. And as a matter of fact the Man I spoke with was Taylor. And I think his first name was Brian.And we only spoke on the phone, for a long time though. He called me at work. And that was in the same town of Envil where I mailed the First Check to KT with the Drawings.
    And the Gazette picture was of the Jet Ski Fender. And the Solid Swim Platform was in the Drawing in the Gazette.I had a copy of it. I still may.
    All I want is a Lawyer, and My Freedom. and the right to collect damages. And then the right to Patent again. Please stop playing games with my life. Constantly going back and forth on whether I have it all figured out or not. I know HRB was punished when no one else was. I also know that some where along the terrible line Someone should have told me. but they didn’t and then proceeded to waste 15 plus years of my life. Take my husbands Job from him after they got him up here. Destroy my Computers my Evidence and then worm me. And if you want more, I’ll be happy to provide.
    After all I have so much more to share. I want my Freedom. This Fender does not fit into DOD! And I do not belong in chains.

  6. 16

    As I recall, one of the fundamental questions one asks when drafting claims is: “Who is the infringer.” And the way I was taught is that if a single entity with $$$ doesn’t infringe the claim, try again. And again. And if there are problems getting it to happen, the client needs to be made aware of that.

    Going back to the Prometheus case, would a patient who asks a lab to test his/her thioguanine levels and then provides the information to a doctor be guilty of joint infringement if the doctor advises a change in the level of drug administered? Judge Lourie says the doctor alone isn’t infringing any of the claims. Obviously the lab is not infringing (the test was in the prior art). Are the doctor and the lab both “agents” of the patient? Does the doctor’s ethical obligation to provide sound advice to the patient render the so-called “warning step” (in quotes because it appears nowhere in the claims) a non-optional step?

  7. 15

    I personally found many practitioners will be unaware of the issue, even in the last couple of years. I was somewhat aware of the issue only because I was involved in the case where the issue came up. This was back in the 90s. My client was accused of infringement of claims directed to a sputtering machine. Both apparatus and method claims not only included the sputtering machine and its structure, but also required apparatus for texturing the discs on which the sputtering machine operated. The problem for the patent owner was that the texturing apparatus and steps were provided by independent contractors.

    The fact that we were sued at all illustrates that the whole issue of lack of direct infringement was not well understood. If I recall correctly, the patent owner’s theory ranged from inducement, to contributory, to joint liability. They really had a problem.

    But what I found over the years is that this issue was not well understood in the patent bar. Many, many patents – even patents being drafted today, have this problem. Many, if not most practitioners, simply do not understand the issue.

    So whose fault is that? We practitioners, or a Federal Circuit that until recently never made the issue clear.

  8. 14

    I suspect BMC caught a lot of practicioners off guard. Fess up. How many here would have drafted method claims with multiple actors thinking there would be joint infringement and joint and several liabiltiy?

    I read the IPO brief in Muniauction. It said that that BMC caught everyone off guard and that that case would sink a lot of patents – unexpectedly.

  9. 12

    the customers knowingly became agents of Limelight by obliging themselves to perform the missing steps and arrive at an infringing operational method.

    My understanding is that nobody was obliged to perform the tagging step, but that the tagging was merely an optional step that would entitle the customers to some benefit or service. Also, the tagging and resulting hosting process aren’t actually required, because the prior art shows an alternative way of hosting websites without it.

    It’s kind of like if someone had a method of operating a retail establishment with steps for stocking the shelves and operating the cash register and whatnot, and one of the steps was “walking into the store with the intention of buying something”, and you (the accused infringer) opened a shop.

    Now, you don’t perform the “walking” step yourself, so you don’t directly infringe. You’re not forcing or requiring anybody to walk in with the intention of buying something, even though you’re clearly open to the prospect and it’s clearly a condition for them to enter into a contract with you. The people who walk in aren’t your agents or anything, nor are they direct infringers because they don’t do any of the other stuff in the claim.

    So where’s the infringement? Whom to sue? Would it make a difference if you were pretty sure that most people would walk into your shop? That everybody who passed by would walk into your shop? Would it make a difference if you offered free candy inside?

  10. 11

    Netflix is in a contractual relationship with Limelight to help it deliver content faster. Netflix is a commercial concern that in turn produces information demanded by third parties (its customers).

    Wouldn’t Netflix in this case be the principle and Limelight be the sub contractor under direct control of the principle?

  11. 10

    The CAFC holding appears to be bad law.

    It would seem that the “contractual” agreement between Limelight and its customers would have to be the purchase order or supply/use agreement/contract for the services themselves. With both parties knowing that the operational method requires specific steps in order to be operative, with Limelight only providing part of those steps and relying upon its customers to conduct the “missing steps”, and with Limelight enabling its customers to perform the missing steps in order to arrive at an operational method, the customers knowingly became agents of Limelight by obliging themselves to perform the missing steps and arrive at an infringing operational method. By purchasing not-fully-operational TAILORED versions of the software from Limelight and by agreeing to perform the missing steps, the customers acted as agents for Limelight. In other words, without the purposeful and willful cooperative action of both parties (Limelight knowing the customers’ needs and providing tailored versions of incompletely operable software and the customers agreeing to perform certain steps in order to arrive at completely operable software), the customers would not have arrived at the infringing operational method (software) themselves. And clearly there was no “if” about whether or not the customers intended to use Limelight’s software.

  12. 9

    I’m with Courtenay here, fascinating though the theoretical discussion is between Malcolm and Paul.

    The real practical problem, in these cases, is that the claim recites some steps that are not performed by the person you want to sue.

    In this case, you’d have to amend the “tagging” step to something like “receiving information identifying which embedded objects were tagged”, which would not be easily done in a narrowing reissue.

  13. 8

    BTW, as I recall that there are two CAFC decisions that if a more than two year late reissue is found to have improperly broadened a claim the other claims go down the tube as well.

  14. 7

    Thanks “whatever” I think you will also find that most NPE’s do not get their hands on their asserted patents [which they rarely obtained themselves] (by bankruptcies or whatever) until long after those patents are also more than two years old.
    Malcom, re your hypothetical of an original claim to “a method of achieving X, wherein A carries out step (1) and B carries out step (2)” versus a reissue claim to “a method of achieving X, wherein A carries out step (1) and B carries out step (2), wherein A and B are the same entity.”
    That might slip by some examiner, but that IS a claim broadening, because the original claim did NOT cover A and B being the same entity, unless the spec and file history said so, and if not that reissue claim should also get a 112 rejection.

  15. 6

    Phillies, my first thought when reading this opinion was “Netscape? Does anyone still use Netscape??”

    Mooney, I don’t think your suggested reissue claims would solve the problem. Here, the problem was that A and B were not the same entity. What you would want to do is try to write the claim with fewer method steps, to recite only the steps performed by the provider, not the customer. This can be easier said than done . . .

  16. 4

    From the decision:

    “A user requesting a web page using a web browser (e.g., Netscape Navigator® or Microsoft Internet Explorer®) will…”

    Netscape Navigator? Why not use Mosaic?

    And what would happen if I had:

    1. A process comprising steps A and B.
    2. The process of claim 1, wherein steps A and B are carried out by different entities.
    3. The process of claim 2, wherein the entities are unrelated.

    Do I still lose, or have I circumvented the CAFC’s case law?

  17. 3

    PM Is not a reissue that makes anyone an infringer who was not an infringer before a “broadening” reissue that can only be filed within two years of the issue date of the patent?

    As a matter of definition, I don’t think so.

    Claim 1: A method of achieving X, wherein A carries out step (1) and B carries out step (2).

    Reissue claim: “A method of achieving X, wherein A carries out step (1) and B carries out step (2), wherein A and B are the same entity.”

    Is the reissue claim broadening?

  18. 2

    Most of the patent holders do not examine their patent qualities until the time when a potential suit is to be filed. At this timing, the two year window for enlarging the protection scope has already lapsed. However, for NPE, either the protection scope is immediately examined and corrected if necessary, or the prosecution history is carefully processed, from my experiences, not many companies seek reexamination to broaden the protection scope.

  19. 1

    Glad to see the expression “joint infringement” being further clairified as normally an oxymoron.
    However, re: “..the court suggested (citing Mark Lemley’s Divided Infringement article), that patentees may be able to correct a claim by seeking a reissue patent.”
    Is not a reissue that makes anyone an infringer who was not an infringer before a “broadening” reissue that can only be filed within two years of the issue date of the patent? [What percentage of patents are being thoroughly reviewed for their litigation potential at that early date?]

Comments are closed.