April 9, 2026 marked the most severe governance crisis in the history of decentralized AI network Bittensor. Covenant AI, the core subnet developer, announced its exit from the ecosystem and publicly accused project co-founder Jacob Steeves of exercising "centralized control" over the network. Following the news, Bittensor’s native token TAO plummeted from $337 to around $250 within six hours—a 25% drop that wiped out approximately $650 million in market capitalization and triggered about $9.1 million in forced liquidations of long positions. As of April 15, 2026, TAO remained near $250, with market confidence significantly shaken.
Why Did the Core Developer Accuse Bittensor’s Decentralization of Being a Charade?
Covenant AI founder Sam Dare stated bluntly in a public declaration that Bittensor’s much-touted "three-signature governance" amounted to a "decentralization theater." He alleged that Jacob Steeves effectively controlled the three-member ruling team, could bypass consensus to enact unilateral network changes at will, and that the other members served primarily as legal shields.
Specific accusations included Steeves suspending token emissions for Covenant AI’s subnet, stripping the team of management rights over its community channels, publicly deprecating subnet infrastructure without following established procedures, and exerting financial pressure through large-scale token sales during the conflict. Dare’s statement was unequivocal: "Bittensor’s entire core promise—that no single entity can control it—is a lie."
It’s important to note that Covenant AI was not a fringe player in the ecosystem. The team previously completed Covenant-72B—a decentralized large language model pretraining project with 72 billion parameters, collaboratively built by more than 70 independent contributors on general-purpose hardware. The project earned public recognition from NVIDIA’s CEO and was cited by Anthropic’s co-founder. The departure of such a leading developer dealt a substantial blow to Bittensor’s credibility.
How Did the Co-Founder Respond to Centralization Allegations and Attempt to Rebuild Trust?
Jacob Steeves did not deny the existence of the "three-member governance" structure. He acknowledged that Bittensor’s governance documents describe a transitional model in which a "three-member ruling council" composed of Opentensor Foundation employees shares root privileges with the Senate.
Steeves explained that when dTAO launched a year earlier, the team had planned to deploy a community-led subnet governance mechanism—allowing Subnet Alpha holders to elect the hyperparameter team via wallet voting. However, this feature was postponed to grant subnet owners more control in the early stages. On Discord, Steeves stated that now was the right time to revisit this discussion, suggesting the introduction of community voting to elect teams and restart subnets.
At the same time, Steeves sharply criticized Covenant AI’s actions, claiming that Sam Dare’s decisions were "clearly motivated by malice and greed," and that the exit was designed to inflict "maximum pain." He shifted the focus of the controversy from governance structure to personal accountability.
The Market Mechanism Behind TAO’s 25% Drop in Six Hours
The price crash was not caused solely by governance disputes. Market data shows TAO fell from $337 to $253 in six hours, erasing more than $650 million in market value. However, the deeper market impact came from a chain reaction.
About six hours after announcing his departure, Covenant AI founder Sam Dare sold roughly 37,000 TAO tokens from his wallet, worth about $10 million. Analyst Michaël van de Poppe pointed out that the true destructive force was not the governance accusations themselves, but the panic selling and leveraged liquidations triggered by this dump.
The selling pressure set off a negative spiral: validator staking declined, reducing consensus weight; lower rewards weakened incentives; shaken confidence accelerated user attrition. Trading volume surged to $1.72 billion on April 10, a dramatic spike compared to the monthly average of about $500 million.
Can the Lock-Up Staking Proposal Prevent Similar Conflicts Mechanistically?
In response to the controversy, Steeves introduced a new protocol feature called "lock-up staking," which some observers have described as one of the most significant proposals in Bittensor’s history.
The core logic shifts governance foundations from interpersonal trust to cryptographic commitments enforced by on-chain code. Subnet owners must lock their tokens for a fixed period, during which the tokens cannot be transferred, providing the community with a verifiable signal of long-term commitment. Lock-up staking operates on the principle of "time plus stake equals trust"—the product of holdings and remaining lock-up time determines the strength of commitment. The larger the stake and the longer the lock-up, the greater the ownership weight.
Steeves admitted his real mistake was not implementing this supplemental mechanism sooner, believing that if lock-up staking had been launched earlier, it might have prevented the conflict from escalating. If adopted, the proposal would provide subnet owners with an on-chain, auditable guarantee of long-term commitment, helping prevent sudden exits and large-scale token dumps like Covenant AI’s.
However, lock-up staking alone cannot resolve the governance power distribution issues inherent in the "three-member council" structure. It is more a fix for economic incentives than a fundamental overhaul of power dynamics.
What Structural Challenges Does Bittensor’s Valuation Logic Face?
The governance dispute has put Bittensor’s valuation model under intense scrutiny. As of March 2026, Bittensor’s circulating market cap ranged from $2.6 billion to $3.6 billion, with fully diluted valuation between $5.8 billion and $7 billion. Yet the network’s actual external revenue—cash flow generated from users paying for AI models and services—was far below its market valuation.
Take the largest subnet, Subnet 3 (Templar), for example. This subnet receives about $52 million annually in TAO inflation subsidies from the protocol, but its actual external revenue is only about $2.4 million—a subsidy more than 20 times larger than real income. This means the protocol’s valuation is currently heavily dependent on an internally driven economic loop powered by inflation, rather than genuine external business value capture.
The eruption of governance conflict further exposed this structural vulnerability: when a network’s trust foundation is built on market consensus around the "decentralization" narrative, and that narrative is publicly challenged by core builders, the valuation premium faces the risk of repricing.
How Should Governance Power Be Allocated in Decentralized AI Networks?
The central question raised by the Bittensor incident is: in an AI network that claims to be "permissionless and decentralized," how should governance power be distributed?
Covenant AI’s accusations reveal a deeper contradiction: transitional governance structures may boost decision-making efficiency in a project’s early stages, but over time, the tension between concentrated power and decentralization narrative intensifies as the ecosystem grows. Steeves acknowledged the existence of the three-member council but argued it was a temporary arrangement on the path to fully open governance.
The crux of the controversy is not whether a transition period should exist, but who decides its duration, who controls the pace of power decentralization, and whether a fair arbitration mechanism exists when builders and core teams disagree. These questions impact not only Bittensor’s future, but also represent governance challenges common to the entire decentralized AI sector.
Conclusion
The real value of Bittensor’s governance crisis lies not in the price swings, but in exposing the gap between decentralized AI narrative and actual practice. Covenant AI’s accusations thrust the "three-member council" structure into the public eye, while TAO’s 25% crash reflected the market’s sensitivity to narrative credibility.
Mechanistically, the lock-up staking proposal represents a shift from legal accountability to cryptographic enforcement in governance, but it cannot solve fundamental power distribution issues. From a valuation perspective, Bittensor remains in a subsidy-driven phase, with a significant gap between external real revenue and market cap; governance disputes may accelerate market reassessment of this structural disconnect.
For the decentralized AI sector, the Bittensor episode offers a key lesson: decentralization is not a one-time declaration, but an ongoing institutional practice that requires constant validation. The network’s governance structure must match its narrative promises, or even the best builders will eventually vote with their feet.
Frequently Asked Questions (FAQ)
Q1: Why did Bittensor’s TAO token drop sharply on April 10?
The decline was triggered by multiple factors: Covenant AI announced its exit and accused Bittensor of centralized governance, sparking market panic. Its founder then sold about 37,000 TAO (worth roughly $10 million), intensifying selling pressure and triggering about $9.1 million in forced liquidations of long positions. TAO dropped from $337 to $253 in six hours, a fall of about 25%.
Q2: What is the "three-member council" structure?
Bittensor’s governance documents describe a transitional model in which a "three-member council" composed of Opentensor Foundation employees shares network root privileges with the Senate. Covenant AI alleges that in practice, Jacob Steeves effectively controls this structure and can unilaterally implement changes by bypassing consensus.
Q3: How does the lock-up staking mechanism work?
Lock-up staking is a new protocol feature proposed by Bittensor’s co-founder in response to the crisis. Subnet owners must lock their TAO tokens for a fixed period, during which the tokens cannot be transferred. The product of holdings and remaining lock-up time determines the strength of commitment, measuring the team’s long-term dedication to the network. The mechanism aims to replace personal credit with on-chain code, preventing large-scale token dumps or sudden exits.
Q4: What impact does this event have on the decentralized AI sector?
The incident exposed common governance challenges for decentralized AI projects: how to balance early-stage decision efficiency with long-term power decentralization, and how to ensure consistency between "decentralization" promises and actual governance practices. The success or failure of the lock-up staking proposal may offer important mechanistic reference for the industry, but fundamental governance power allocation issues remain to be explored collectively.
