The agreement mandates binding arbitration, waives both jury trials and class actions, and requires disputes to be resolved in San Francisco, California. These provisions significantly restrict access to justice and favor the company over remote parties.
Scope is clearly limited to educational institutions and teachers, explicitly excluding consumer use. Acceptance occurs via electronic consent or service access, lacking an explicit clickwrap mechanism.
Pricing changes require 30 days notice before taking effect, providing reasonable advance warning. The document lacks explicit details on auto-renewal, trial conversions, or proration refunds, reflecting its enterprise API billing model rather than consumer subscriptions.
Anthropic reserves the right to terminate access at its sole discretion for perceived misuse without a defined cure period or appeal process. While convenience termination requires 30 days notice, the discretionary suspension clause creates significant account instability.
Material changes require 30 days advance notice and apply prospectively, preventing retroactive penalties. Regulatory updates bypass the notice period, and the contract lacks an explicit right to cancel with a pro-rated refund upon objection to changes.
Indemnification is reciprocal and fault-based, covering IP infringement for authorized use and claims arising from policy violations or unauthorized inputs. It reasonably excludes fraud and willful misconduct, though labeling it the sole remedy for third-party claims is somewhat restrictive.
Liability is capped at the greater of $100 or prior 12-month fees, and the cap explicitly applies to negligence and tort claims without carving out gross negligence or willful misconduct. While standard for B2B contracts, the broad waiver reduces practical remedies for service failures.
The agreement incorporates a Data Processing Addendum, aligns with FERPA, designates Anthropic as a School Official, and includes robust confidentiality and data destruction obligations. It lacks granular consumer-style opt-outs but provides strong institutional data controls.
Users retain full ownership of inputs and outputs, with Anthropic explicitly disclaiming any rights and prohibiting model training on customer content. This represents industry-leading protection for user-generated and submitted data.