Snapshot 254059
Parsed
What we read from this document, as JSON. Each entry names the snapshot it was read from.
{
"entries": [
{
"kind": "subprocessor",
"vendor": "amazon-web-services",
"evidence_id": 236583,
"data": {
"name": "Amazon Web Services",
"domain": "aws.amazon.com",
"purposes": [
"AWS hosts failover and additional cloud services in parallel with GCP.",
"Cloud provider"
],
"locations": [
"Global"
]
}
},
{
"kind": "subprocessor",
"vendor": "anthropic",
"evidence_id": 236583,
"data": {
"name": "Anthropic",
"domain": "anthropic.com",
"purposes": [
"Anthropic provides LLM inference using proprietary models on a ZDR basis with no modal training.",
"LLM model and inference"
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "assemblyai",
"evidence_id": 236583,
"data": {
"name": "AssemblyAI",
"domain": "assemblyai.com",
"purposes": [
"AssemblyAI transcribes live customer call audio, operating as the failover and quality-comparison provider alongside our primary transcription vendor.",
"Speech-to-text transcription for voice calls"
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "baseten",
"evidence_id": 236583,
"data": {
"name": "Baseten",
"domain": "baseten.co",
"purposes": [
"Baseten provides LLM inference using open-weight models on a ZDR basis with no modal training.",
"LLM inference"
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "braintrust",
"evidence_id": 236583,
"data": {
"name": "Braintrust",
"domain": "braintrust.dev",
"locations": [
"Testing and monitoring of low-level LLM tasks"
]
}
},
{
"kind": "subprocessor",
"vendor": "cartesia",
"evidence_id": 236583,
"data": {
"name": "Cartesia",
"domain": "cartesia.ai",
"purposes": [
"Text to speech"
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"evidence_id": 236583,
"data": {
"name": "Celeris",
"domain": "celeris.ai",
"purposes": [
"LLM inference",
"LLM inference on a ZDR basis"
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "cerebras",
"evidence_id": 236583,
"data": {
"name": "Cerebras Systems",
"domain": "cerebras.ai",
"locations": [
"LLM inference provider"
]
}
},
{
"kind": "subprocessor",
"vendor": "deepgram",
"evidence_id": 236583,
"data": {
"name": "Deepgram",
"domain": "deepgram.com",
"purposes": [
"Deepgram converts live and recorded customer call audio into text so our AI voice agents can understand what a caller said, and to produce post-call transcripts.",
"Speech-to-text transcription for voice calls"
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "digitalocean",
"evidence_id": 236583,
"data": {
"name": "DigitalOcean",
"domain": "digitalocean.com",
"purposes": [
"DigitalOcean does not store inputs or outputs on its infrastructure for the open-weight models we use, and does not use customer data to train models.",
"LLM inference capacity for open-weight models"
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "elevenlabs",
"evidence_id": 236583,
"data": {
"name": "ElevenLabs",
"domain": "elevenlabs.io",
"purposes": [
"ElevenLabs synthesises the speech our AI voice agents use on calls, and creates the custom voices subscribers select.",
"Text-to-speech voice synthesis"
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "fireworks",
"evidence_id": 236583,
"data": {
"name": "Fireworks AI",
"domain": "fireworks.ai",
"purposes": [
"LLM inference",
"Provides LLM inference services for our AI agent system."
],
"locations": [
"United States"
]
}
},
{
"kind": "subprocessor",
"vendor": "google-cloud",
"evidence_id": 236583,
"data": {
"name": "Google Cloud Platform",
"domain": "cloud.google.com",
"locations": [
"Cloud provider"
]
}
},
{
"kind": "subprocessor",
"vendor": "groq",
"evidence_id": 236583,
"data": {
"name": "Groq",
"domain": "groq.com",
"purposes": [
"LLM inference"
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "livekit",
"evidence_id": 236583,
"data": {
"name": "LiveKit",
"domain": "livekit.io",
"purposes": [
"LiveKit Cloud carries the real-time audio of voice calls between the telephone network and our AI voice agents, and composes the call recording that is written to our own storage.",
"Real-time media transport for voice calls"
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "merge-api",
"evidence_id": 236583,
"data": {
"name": "Merge",
"domain": "merge.dev",
"purposes": [
"Integration with third-party ticketing systems"
],
"locations": [
"San Francisco, California"
]
}
},
{
"kind": "subprocessor",
"vendor": "modal",
"evidence_id": 236583,
"data": {
"name": "Modal",
"domain": "modal.com",
"purposes": [
"LLM inference services",
"Modal provides LLM inference using open-weight models on a ZDR basis with no modal training."
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "openai",
"evidence_id": 236583,
"data": {
"name": "OpenAI",
"domain": "openai.com",
"purposes": [
"LLM model and inference",
"OpenAI provides LLM inference using proprietary models on a ZDR basis with no modal training."
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "openrouter",
"evidence_id": 236583,
"data": {
"name": "OpenRouter",
"domain": "openrouter.ai",
"purposes": [
"Provides LLM inference services on a ZDR and no-model-training basis."
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "twilio",
"evidence_id": 236583,
"data": {
"name": "Twilio",
"domain": "twilio.com",
"purposes": [
"Telephony and messaging connectivity",
"Twilio provides the phone numbers and carrier connectivity that carry voice calls, SMS and WhatsApp messages between end customers and our AI agents."
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "typesafe-ai",
"evidence_id": 236583,
"data": {
"name": "TypeSafe AI",
"domain": "typesafe.ai",
"purposes": [
"LLM inference services",
"LLM inference services provided on a ZDR basis."
],
"locations": [
"USA"
]
}
},
{
"kind": "subprocessor",
"vendor": "vercel",
"evidence_id": 236583,
"data": {
"name": "Vercel",
"domain": "vercel.com",
"purposes": [
"AI Gateway services — a unified API routing layer that load-balances LLM inference requests across our existing providers for improved reliability and performance.",
"Vercel's AI Gateway operates a Zero Data Retention (ZDR) policy — prompts, outputs, and sensitive data are not retained after requests are completed."
],
"locations": [
"USA"
]
}
}
]
}