Terms of Service Change History

The new version entirely removes Section 1.3 which explicitly stated 'ANYSPHERE WILL NOT USE CONTENT TO TRAIN, OR ALLOW ANY THIRD PARTY TO TRAIN, ANY AI MODELS, UNLESS YOU'VE EXPLICITLY AGREED.' This was a core user protection regarding data usage for model training. The new version contains no equivalent safeguard or model training policy.

Effective: October 1, 2026

The new version removes Section 1.5 (Use Restrictions) which prohibited reverse engineering, model extraction attacks, competitive model development, data scraping, and other critical protections. It also removes Section 1.4 acknowledging AI model limitations and user responsibility for Suggestion accuracy. These sections provided important legal guardrails for users.

Effective: October 1, 2026

The original ToS is replaced with a French-language pricing policy that includes Section 3.2 stating marketing rights are 'taken into account in the calculation of our fees' with limited opt-out only if 'expressly indicated' in purchase order. The original version provided clearer user control; this version presumes consent and ties it to pricing.

Effective: October 1, 2026

The new version removes Section 1.6 which contained 'ANYSPHERE SHALL HAVE NO LIABILITY WHATSOEVER ARISING OUT OF OR IN CONNECTION WITH BETA SERVICES.' This liability limitation for beta/experimental features is entirely absent from the new document.

Effective: October 1, 2026

The new version introduces detailed pricing sections (Sections 1-8) establishing pre-engaged usage commitments, on-demand overages, true-up periods with 30-day active user calculations, token-based billing for models, and automatic price adjustments per catalog rates. The original version contained no pricing terms, making this a substantial new commercial framework.

Effective: October 1, 2026