NEAR AI said its Cloud API is now OpenAI-compatible, allowing developers to point existing clients at a new base URL while keeping the rest of their code unchanged. The company also said requests are proxied through NEAR AI, so the model provider sees NEAR AI as the source rather than the end user or the workload owner.

The launch was outlined in NEAR AI’s announcement post, which describes the API as private-chat focused and says the privacy level depends on the model selected. According to the post, third-party models are proxied through NEAR AI, while models marked TEE are designed to keep prompts and outputs unreadable to anyone else, including NEAR.

Compatibility is the main change; privacy depends on the route

For developers, the practical change is that an OpenAI-style integration can be redirected to NEAR AI without rewriting the application logic. NEAR AI said the same API surface can be used while the underlying model choice determines how the request is handled and how much of the data remains visible outside the secure environment.

The company’s description suggests two layers of privacy handling. In the proxy setup, the lab receives the request from NEAR AI rather than from the original user. In the TEE path, NEAR AI said the prompt and output remain unreadable to anyone else, including NEAR itself.

The announcement does not change the basic fact that this is still a cloud API service, but it does position the product as a compatibility layer for teams that want to move existing OpenAI-based workflows without changing the client code structure.

The post NEAR AI Cloud API Adds OpenAI Compatibility and Tiered Privacy appeared first on The Cryptocurrency Post.

Aptos Labs said research on a quantum-resistant Byzantine fault tolerant, or BFT, consensus design has been accepted to IEEE S&P ’27. In the company’s announcement on X, Aptos Labs said the paper shows BFT consensus can be made quantum-resistant without sacrificing performance.

Congratulations to Aptos Labs’ @XiangZhuolun and his co-authors: Simple-IT has been accepted to IEEE S&P ’27! 🎉

The paper shows that BFT consensus can be made quantum-resistant without trading away performance.

Learn more below. https://t.co/ihkyhwSVaV pic.twitter.com/bO917aTE6a

— Aptos Labs (@AptosLabs) September 15, 2026

The update centers on a research result rather than a network change. Aptos did not say in the announcement that the paper represents a live protocol upgrade, only that the work has been accepted for presentation at the 2027 edition of IEEE’s security and privacy conference.

An additional post from Aptos co-founder Avery Ching congratulated researcher Xiang Zhuolun and co-authors, saying the paper, Simple-IT, had been accepted to IEEE S&P ’27. That post echoed the same core claim: that the design can preserve performance while adding quantum resistance.

Research claim, not deployment status

Based on the company’s wording, the key point for readers is that Aptos is highlighting a research contribution in consensus design, not announcing that users or validators must change anything immediately. The announcement frames the work as evidence that post-quantum security can be explored without an obvious performance trade-off.

That distinction matters because acceptance at a conference signals academic recognition of the paper, while actual network behavior would depend on whether and when any of the ideas are implemented in production.

The company did not provide implementation timing, rollout details or network-wide activation plans in the announcement. For now, the development is best read as a technical milestone for Aptos Labs’ research team.

The post Aptos research on quantum-resistant BFT consensus accepted to IEEE S&P ’27 appeared first on The Cryptocurrency Post.

NEAR AI said its cloud offering now supports an integration with Intel Trust Authority that it describes as advancing confidential inference and independent attestation. In a separate public post, the NEAR Protocol account said more than 500,000 NEAR has been staked toward confidential inference on NEAR AI Cloud, with more than 40 models supported, including those from Anthropic, OpenAI and Google.

The official blog post frames the integration as a step toward confidential AI execution, where inference can be performed in a protected environment. The announcement does not explicitly use the term “formal verification,” even though that phrase has appeared in some references to the story.

What NEAR is signaling around its AI cloud

The clearest confirmed detail is that NEAR AI is linking its cloud product with Intel Trust Authority and presenting the setup as a way to support confidential inference with attestation and verification features. That makes the update more specific than a broad AI branding push: it points to a technical change in how the cloud service is being positioned and operated.

The staking figure shared by NEAR Protocol suggests active participation around that cloud product, but the post itself does not spell out the mechanics behind the staking, the incentives involved or whether the figure reflects user demand, ecosystem support or another form of commitment. The company’s message does confirm the scale it wants readers to notice: 500,000+ NEAR and 40+ supported models.

For readers tracking AI-focused crypto projects, the relevant distinction is between a product that is being described as confidential and one that is broadly live and verified at the application layer. NEAR’s announcement supports the first point clearly. The broader operational and adoption picture remains limited to what the company has publicly said.

The post NEAR AI says confidential inference is expanding across its cloud stack appeared first on The Cryptocurrency Post.