The legal dispute between Rippling and AI infrastructure startup Runlayer has entered a new phase, transforming a failed enterprise software trial into a broader confrontation over intellectual property, product development and the competitive power of large technology companies.

Rippling filed a lawsuit on August 10 accusing Runlayer of infringing three of its patents. The move comes roughly two weeks after Runlayer sued Rippling, alleging breach of contract and misuse of its intellectual property after Rippling evaluated Runlayer’s technology and later developed its own competing MCP gateway.

The dispute is significant because it illustrates a growing structural problem in the AI software market: startups can build highly specialized infrastructure, but their largest potential customers may possess enough engineering capacity to reproduce similar capabilities internally.

Runlayer developed a secure gateway based on the Model Context Protocol, or MCP, an open standard that enables AI models and agents to connect with external data and software systems. The company combines that infrastructure with cybersecurity capabilities such as threat detection and controls designed to govern how AI agents interact with enterprise resources.

The product sits within one of the fastest-growing layers of the AI technology stack.

As enterprises deploy more AI agents, they need mechanisms to determine which systems those agents can access, what information they can retrieve and what actions they can perform.

That creates a market for infrastructure sitting between AI agents and corporate systems.

MCP gateways are designed to provide precisely that control layer.

For Runlayer, the opportunity was to become an infrastructure provider at a time when companies are rapidly experimenting with autonomous AI systems.

The company, founded by Andrew Berman, has raised $42 million from investors including Khosla Ventures and Felicis. Its previous experience in enterprise-oriented technology also gave it a foundation for selling a complex security product to large organizations.

Rippling became one of its early prospective customers.

The relationship reportedly lasted nearly a year and involved intensive technical evaluation. Runlayer says it shared sensitive information during the process, including elements of its product roadmap and source code, under confidentiality and trial agreements designed to restrict copying and derivative development.

But the commercial relationship never reached a paid agreement.

The two companies could not agree on pricing, and the trial eventually ended.

That is where the business relationship turned into a legal confrontation.

Runlayer later alleged that Rippling had used information obtained during the evaluation to develop what it described as a near-identical product.

Rippling has denied the allegations, saying it developed its own MCP gateway using proprietary information and has characterized Runlayer's lawsuit as an attempt to avoid competition.

Rippling's countersuit changes the balance of the dispute.

Rather than simply defending itself against accusations of trade-secret misuse, the company is now asserting that Runlayer itself infringed three Rippling patents.

The courts will ultimately have to determine whether either company's claims have legal merit.

But the strategic implications are already visible.

The dispute demonstrates how the economics of AI enterprise software differ from those of traditional software markets.

A startup can spend years developing a specialized product and then face a difficult question when it approaches a large potential customer: is the customer going to buy the technology, or use the evaluation process to determine whether it can build an internal alternative?

That question becomes particularly important in AI because software development is becoming faster.

Large companies increasingly have access to powerful foundation models, coding agents and internal engineering teams capable of turning ideas into functional products in relatively short periods.

The traditional distinction between software buyer and software developer is therefore becoming less clear.

A company that previously needed to purchase specialized infrastructure may now be capable of reproducing a significant portion of that infrastructure itself.

This creates a serious strategic challenge for startups.

Enterprise sales can require extensive technical access.

Customers often need to test software against their own systems before signing contracts, particularly when the product controls sensitive data or critical workflows.

But every additional layer of access can increase the risk that the customer learns enough about the technology to build an alternative.

The Runlayer-Rippling dispute is therefore a warning for both sides of the enterprise relationship.

For startups, protecting intellectual property is not simply a legal exercise.

It becomes part of product strategy.

Founders need to determine how much technical information to expose during trials, which components should remain isolated, how source code and architecture are protected, and what contractual restrictions are necessary before providing deep access.

For enterprise buyers, the issue is equally important.

A company may have legitimate reasons to build internally after evaluating a vendor product, particularly when the vendor's pricing, performance or strategic fit does not meet its needs.

But the line between independently developing a competing product and using confidential information obtained during a trial can become legally and commercially sensitive.

That boundary becomes even harder to define when the underlying technology is based on an open standard.

MCP itself is open.

Companies are therefore free to develop products that use the standard.

The competitive question is not necessarily whether Rippling can build an MCP gateway.

It can.

The harder question is whether the specific implementation incorporates protected intellectual property belonging to another company.

That distinction will likely be central to the legal battle.

The case also illustrates the increasing importance of MCP in the enterprise AI ecosystem.

Originally introduced as an open protocol for connecting AI models to external tools and data, MCP has evolved into a foundational interoperability layer for agentic systems.

As enterprises deploy AI agents across more applications, controlling these connections becomes increasingly valuable.

A gateway can function as a policy and security layer between an agent and the systems it wants to access.

This gives companies the ability to monitor requests, enforce permissions and potentially detect suspicious behavior.

The commercial opportunity is consequently larger than simply providing connectivity.

The winning infrastructure providers may become the control points through which enterprises manage their AI agents.

That makes the market strategically important.

It also explains why established software companies may be reluctant to depend entirely on startups for such a critical layer.

For Rippling, developing its own gateway could offer several advantages.

The company can integrate the technology directly into its existing software platform, customize it for its own customers and control the product roadmap.

It can also avoid becoming dependent on an external infrastructure provider for a capability that could eventually become central to its AI strategy.

This reflects a broader pattern in enterprise technology.

Large software companies frequently turn internal tools into commercial products when they become strategically important.

Rippling has already followed a similar approach with other internally developed technologies, including its AI Spend Console.

The strategy can be economically attractive.

Instead of paying a third-party vendor indefinitely, a company can invest engineering resources once and potentially create a proprietary capability that strengthens its broader platform.

For startups, however, this creates a difficult market dynamic.

The better their technology becomes, the more attractive it may be for major customers to replicate or replace it.

That makes differentiation critical.

A startup cannot rely solely on the fact that it built a particular feature first.

It needs advantages that are difficult to reproduce, whether through proprietary technology, specialized security capabilities, accumulated data, operational expertise, network effects or customer relationships.

The Runlayer case also highlights the vulnerability of companies operating in crowded infrastructure markets.

MCP has attracted growing competition because it addresses a fundamental requirement of agentic AI: enabling models to interact with external systems safely.

As the standard becomes more widely adopted, infrastructure around it is likely to become increasingly commoditized.

That creates pressure on startups to move beyond basic protocol support and build higher-value services around governance, security, observability and enterprise management.

Runlayer's combination of an MCP gateway with security capabilities reflects that strategy.

But the existence of a large enterprise competitor developing its own product demonstrates how difficult it can be to maintain differentiation at the infrastructure layer.

The legal dispute may therefore have consequences beyond the two companies.

If enterprise buyers become concerned that product trials can later generate intellectual-property disputes, they may demand more restrictive testing environments.

Startups may respond by limiting access to source code or sensitive architecture.

Customers, meanwhile, may increasingly prefer vendors with clearer contractual frameworks and stronger technical isolation.

That could make enterprise procurement more complicated.

It could also increase the cost of selling sophisticated AI infrastructure.

The irony is that deep technical integration is often necessary to demonstrate value.

An enterprise customer cannot fully evaluate an AI security product without allowing it to interact with real systems.

But the deeper that integration becomes, the more sensitive the relationship can become.

This tension is likely to grow as AI infrastructure becomes more deeply embedded in corporate operations.

The dispute also reveals how intellectual property law is being tested by the speed of AI development.

Traditional software disputes often revolve around source code, patents or trade secrets.

AI products introduce additional layers.

A system may incorporate open standards, proprietary implementations, model behavior, prompts, orchestration logic and security policies.

Determining which elements constitute protectable intellectual property can be complicated.

Companies therefore have strong incentives to establish clear ownership boundaries before commercial collaboration begins.

For investors, the case offers another lesson.

AI infrastructure startups may appear attractive because they serve a rapidly expanding market.

But the customer concentration risk can be substantial.

If a startup depends on a handful of large enterprises, each customer can simultaneously represent major revenue potential and a potential competitive threat.

A major customer may have enough resources to become a competitor.

That means investors and founders need to evaluate not only market size, but also the ability of customers to internalize the technology.

The issue is particularly relevant in the current AI environment because engineering productivity is changing rapidly.

Coding assistants and AI agents can accelerate software development, reducing the time and cost required to reproduce certain classes of applications.

This does not mean every enterprise can easily recreate sophisticated infrastructure.

Security, reliability, compliance and operational maturity remain difficult.

But the barrier to experimentation is falling.

That makes the startup's unique advantages more important.

The Runlayer-Rippling conflict also illustrates an uncomfortable reality of enterprise sales.

A successful product trial is not necessarily a successful business relationship.

A startup may invest substantial engineering resources helping a potential customer integrate and evaluate its product without receiving a contract.

The customer receives valuable knowledge during the process, while the startup bears much of the cost.

If the deal fails, the startup may be left competing against a product built by the customer.

This dynamic could encourage more AI startups to rethink how they structure pilots.

Shorter trials, limited technical access, stronger contractual protections and clearer commercial commitments could become more common.

Enterprises, meanwhile, may face greater scrutiny over how they use vendor information after a pilot ends.

The outcome of the legal cases will therefore be closely watched by both sides of the AI ecosystem.

For Runlayer, the priority is protecting its intellectual property and preserving its ability to compete in a market where larger companies can potentially replicate infrastructure.

For Rippling, the objective is defending itself against accusations of misappropriation while establishing that its own product development does not violate Runlayer's rights.

Neither side can afford to ignore the commercial consequences.

Even a legal victory may come at a cost if the dispute delays product launches, consumes engineering resources or damages relationships with enterprise customers.

For the broader market, however, the dispute may ultimately serve as a useful warning.

AI has reduced the distance between having an idea and building a product.

That is good news for innovation.

But it also reduces the distance between being a customer and becoming a competitor.

The old enterprise software model assumed that vendors developed products and customers purchased them.

The AI era increasingly allows customers to evaluate, modify and potentially reproduce those products themselves.

That changes the balance of power.

Startups will need stronger defenses around intellectual property and sharper differentiation.

Large companies will need clearer internal governance around information obtained from vendors.

And both sides will need to recognize that AI infrastructure is becoming strategically important enough to justify serious investment.

The legal battle between Rippling and Runlayer may eventually be resolved through settlement or court judgment.

But the business lesson is already visible.

In an AI market where development cycles are accelerating, a startup's biggest prospective customer can also become its most capable potential competitor.

That makes trust, contracts and intellectual-property boundaries as important to enterprise AI strategy as technology itself.