Covenant AI Exits Bittensor, TAO Plummets
Covenant AI left the decentralized Bittensor network over alleged centralized governance, triggering a roughly 15% drop in the TAO token — from about $337 to $284, trading near $292 afterward. Covenant AI said Bittensor’s governance concentrated operational control with a small set of actors and accused founder Jacob Steeves of unilateral actions including suspension of emissions, removal of moderation controls, deprecation of subnet infrastructure, and timed large token sales. Covenant AI will continue developing its Covenant-72B model and related research off-network. Bittensor pushed back, with Steeves saying this episode will yield subnets that run headless and as commodities. The dispute highlights governance fragility in token-linked ML marketplaces and the economic coupling between model performance and token dynamics.
What happened
Covenant AI left the decentralized Bittensor network citing concentrated governance and operational control, precipitating a 15% fall in the TAO token (roughly $337 to $284, later around $292). Covenant AI said it can no longer build on Bittensor after a series of unilateral actions it attributes to founder Jacob Steeves, and will continue its work — including on Covenant-72B — outside the network.
Technical details
Covenant AI claims specific operational measures restricted their subnet: suspension of emissions to their subnet, removal of moderation privileges, deprecation of subnet infrastructure, and large token sales timed with conflicts. Bittensor’s economics link subnet performance and token flows: Subnet 3 and its Covenant-72B model drove a March rally (TAO up ~90%; some subnet tokens up ~400%) by boosting staking-backed automated market maker interactions. That coupling means governance disputes can translate quickly into token-price and liquidity shocks.
- •Tokenomics risk: staking and AMM linkages amplify economic pressure from governance actions
- •Operational risk: centralized control points (emissions, moderation, infra deprecation) undermine permissionless claims
- •Reputation risk: high-profile endorsements (e.g., Jensen Huang, Chamath Palihapitiya) can accelerate inflows, but also concentrate attention and counterparty risk
Context and significance
This is a test case for blockchain-native ML marketplaces where governance, economic incentives, and model development intersect. Decentralized AI depends on credible permissionlessness; when control concentrates, both contributors and capital can exit quickly. For practitioners, the episode surfaces three structural problems: governance design for protocol upgrades and operator powers, economic coupling between token mechanics and model incentives, and the operational tooling needed to isolate subnet control from network-level authority.
What to watch
Will Bittensor revise governance or add on-chain checks to constrain unilateral actions? Will Covenant AI publish migration details and technical choices for off-network training? Market liquidity in TAO and subnet tokens will remain a short-term barometer.
Scoring Rationale
The exit and token drop materially affect participants in Bittensor and signal governance fragility in tokenized ML platforms — important for practitioners building or evaluating decentralized AI. Recent timing reduces peak novelty slightly.
Practice with real FinTech & Trading data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all FinTech & Trading problemsStep-by-step roadmaps from zero to job-ready — curated courses, salary data, and the exact learning order that gets you hired.
