Bittensor’s Decentralized Lie Exposed, Covenant AI Announces the Entire Team Has Left

TAO2,39%

Covenant AI退出Bittensor

Decentralized AI training team Covenant AI announced on April 10 that it is exiting the Bittensor network, and specifically called out the network’s key figure, Jacob Steeves, accusing him of betraying its decentralization promises. Covenant AI founder Sam Dare said: “The entire core commitment of Bittensor—that no single entity can control it—is a lie.”

Accusation List: Five Suppression Actions by Const Against Covenant AI

In its statement, Covenant AI detailed a series of specific actions that Const took against its team, forming the direct basis for its decision to withdraw:

Sam Dare Accuses Const of the Main Suppression Actions

Pausing token emissions: Const unilaterally paused the token emissions for Covenant AI’s subnet under its control, directly cutting off its source of economic incentives within the Bittensor ecosystem

Revoking community administration permissions: Const stripped Covenant AI of control over its own community channels, effectively taking control of its external communication channels

Decommissioning subnet infrastructure: Const unilaterally decommissioned Covenant AI’s subnet infrastructure, causing its technical deployments on the Bittensor network to fall into collapse

Applying pressure through large-scale token selloffs: During the period of conflict between the two sides, Const applied economic pressure on Covenant AI through large-scale, high-visibility token selloffs

Bypassing the consensus mechanism: All of the above actions were not carried out through the network’s formal consensus process for governance, showing effective individual control over the multisig mechanism

Bittensor “Decentralization Theater”: The truth behind the 3-of-multisig of three people

Covenant AI’s core accusations point to the fundamental gap between Bittensor’s claimed governance mechanism and its real-world operation. Bittensor publicly markets that it uses a “3-of-multisig” governance framework as an institutional safeguard for decentralization. However, Covenant AI directly states that Const exerts effective control over this multisig mechanism, enabling unilateral changes to be pushed through at any time while bypassing consensus, and that the other multisig participants are merely “a shield to carry legal responsibility.”

If this allegation is true, it would mean that while Bittensor’s governance architecture is decentralized at the technical design level, it is still led by a single individual at the operational level—forming a kind of “decentralization theater,” where decentralization exists in documents and whitepapers, but centralization exists in real decision-making.

Covenant AI’s confidence: This exit is not a failure—it’s a deliberate choice

The credibility of this withdrawal statement is built on Covenant AI’s actual technical achievements. Covenant-72B is, to date, the largest decentralized LLM pretraining project, with 72 billion parameters, spanning participation from over 70 independent contributors. It received public recognition from NVIDIA CEO, and was cited by Anthropic’s co-founder—making its reputation in the AI industry exceed the Bittensor ecosystem itself.

Covenant AI announced that the team, research outcomes, and models will all be taken along, and also previewed that new project announcements will be released soon, showing that this exit is more like an intentional ecosystem migration rather than a forced departure.

Frequently Asked Questions

What is the core reason Covenant AI is exiting Bittensor?

According to the public statement by Covenant AI founder Sam Dare, the core reason for the exit is that Bittensor’s core figure Const (Jacob Steeves) took a series of suppression actions against its team, including pausing token emissions, revoking community administration permissions, decommissioning infrastructure, and applying economic pressure through token selloffs. Moreover, all actions were carried out unilaterally by bypassing formal governance consensus.

What problems exist with Bittensor’s 3-of-multisig governance?

Covenant AI points out that in real-world operations, Bittensor’s claimed “3-of-multisig” governance framework is under Const’s effective control, leaving other multisig participants without real checks and balances. They only bear legal responsibility and are unable to prevent Const from unilaterally pushing changes. Covenant AI directly describes this as “decentralization theater.”

What is Covenant AI’s Covenant-72B project?

Covenant-72B is Covenant AI’s decentralized LLM pretraining project completed within the Bittensor ecosystem. It has 72 billion parameters, involves over 70 independent contributors, and is the largest of its kind to date. The project received public recognition from NVIDIA’s CEO and was cited by Anthropic’s co-founder, giving Covenant AI substantial industry influence even at the time of its exit.

Disclaimer: The information on this page may come from third parties and does not represent the views or opinions of Gate. The content displayed on this page is for reference only and does not constitute any financial, investment, or legal advice. Gate does not guarantee the accuracy or completeness of the information and shall not be liable for any losses arising from the use of this information. Virtual asset investments carry high risks and are subject to significant price volatility. You may lose all of your invested principal. Please fully understand the relevant risks and make prudent decisions based on your own financial situation and risk tolerance. For details, please refer to Disclaimer.

Related Articles

DeepSeek V4-Flash goes live on Ollama Cloud, US-hosted: Claude Code, OpenClaw one-click integration

Ollama Cloud has launched DeepSeek V4-Flash, with inference hosted on U.S. servers, providing three sets of one-click commands to connect Claude Code, OpenClaw, and Hermes. V4-Flash/V4-Pro use a MoE architecture, with native support for 1M context, and reduce costs with Token-wise compression + DSA sparse attention. In a 1M scenario, token FLOPs per token drop by 27%, and KV cache drops by 10%. API-compatible with OpenAI ChatCompletions and Anthropic, making it easy to switch between multiple workflows and lowering costs and data-sovereignty risk.

ChainNewsAbmedia27m ago

Web3 AI Infrastructure AIW3 Raises $2M in Seed Funding Led by Buffalo Capital

Gate News message, April 24 — Web3 AI infrastructure platform AIW3 announced the completion of a $2 million seed round funding. The round was led by Buffalo Capital, with GalaXin Capital and Three-stones Ventures participating as co-investors. AIW3 is transitioning toward an Agent-as-a-Service

GateNews49m ago

Cohere Acquires German AI Firm Aleph Alpha, Secures $600M Investment for European Expansion

Gate News message, April 24 — Canadian AI company Cohere announced plans to acquire German AI firm Aleph Alpha to strengthen its presence in Europe. Schwarz Group, a backer of Aleph Alpha, plans to invest $600 million in Cohere's Series E funding round. The funding round is expected to close in 202

GateNews1h ago

Xpeng, Xiaomi Lead In-Car AI Push at Beijing Auto Show

Gate News message, April 24 — Chinese automakers showcased advanced in-car AI systems at the Beijing Auto Show on April 24, as the country accelerates its AI Plus strategy and seeks greater independence from foreign semiconductors. Xpeng demonstrated voice-controlled parking that allows drivers to

GateNews2h ago

Former ByteDance Seed Engineer: ByteDance AI Iteration Takes Six Months vs Google's Three Months

Gate News message, April 24 — Zhang Chi, a former engineer at ByteDance's Seed team and current assistant professor at Peking University, revealed on the podcast "Into Asia" that ByteDance requires approximately six months to complete one full cycle of large language model training (pretraining

GateNews2h ago

OpenAI Engineer Clive Chan Challenges V4 Hardware Recommendations, Citing Errors and Vagueness vs. V3

Gate News message, April 24 — OpenAI engineer Clive Chan has raised detailed objections to the hardware recommendations chapter in the V4 technical report, calling it "surprisingly mediocre and error-prone" compared to the acclaimed V3 version. V3's hardware guidance, which included Q&A sessions

GateNews2h ago
Comment
0/400
No comments