What started as a discreet collaboration within a rapidly expanding decentralised AI network has now become a highly visible consequence. When pressure escalates, Covenant AI's exit from Bittensor seems to be more than just another argument; it's a moment that makes people wonder how truly "decentralised" these systems are. A deeper conflict between values and control that the ecosystem can no longer ignore is indicated by the allegations, the counterattacks, and even the market reaction.

• Escalating Governance Tensions Within Bittensor
• Covenant AI's Strategic Role & Sudden Exit
• Token Dump Allegations Spark Fresh Controversy
• $TAO Price Swings Reflect Growing Uncertainity
• Reassessing the Limits of Decentralised Control

Escalating Governance Tensions Within Bittensor

Covenant AI didn't go off lightly; instead, they made specific accusations. Jacob Steeves is at the centre of it all, and the group accuses him of tightening control over areas of the network that were meant to be community-driven and open.

It seems that Covenant's operations-related stoppage of subnet emissions was the tipping point. These were not insignificant elements; rather, they were essential to the team's contributions of compute, training, and fine-tuning capabilities. Losing those emissions meant losing operational continuity as well as influence.

When Covenant apparently lost moderation authority on important subnets, the situation got worse. This was more than simply a technical choice for the crew that contributed to the development and upkeep of those systems; it was like being kicked out of their own workspace. The scenario swiftly changed from disagreement to mistrust when you added the allegation that token sales were used as leverage.

Covenant AI's Strategic Role & Sudden Exit

The involvement of Covenant AI in the ecosystem is what distinguishes this fallout. They built the infrastructure at the heart of Bittensor; they were more than just participants.

One of the network's more ambitious results was a large-scale, permissionless language model trained on common hardware due to their work on pre-training, compute distribution, and fine-tuning. There was recognition for that accomplishment. Industry titans like Jack Clark and Jensen Huang agreed with its importance.

Thus, it creates uncomfortable queries when a contributor of that magnitude leaves. It's not just about governance; it's also about whether or not contributors can actually function autonomously in a purportedly open system.

https://t.co/06wplkr0GY— covenant (@covenant_ai) April 9, 2026