Environmental & Social Responsibility

Quick Summary

TL;DR - What You Need to Know

We use Groq (with a "q"), not xAI's "Grok" - Different companies entirely
No training on your data: Neither Groq nor Anthropic's API use your prompts/outputs to train models by default
Data retention: Anthropic API deletes within 30 days; Groq doesn't permanently retain prompts (short-lived cache only)
Regional hosting: EU customers → Azure EU regions; US customers → Azure US regions for optimal performance
Environmental efficiency: 75% Groq (claims ~10× lower energy than GPUs) + 25% Anthropic (AWS renewable matching) + RAG minimizes tokens
Corporate responsibility: Most spend goes to venture-backed Groq, not Big Tech platforms

Our AI Provider Choices

                    Important: We use Groq (G-R-O-q), the semiconductor + inference company, not xAI's "Grok" chatbot. They're completely different companies.
                

Our Current Mix

75% Groq

25% Anthropic

Groq (G-R-O-q) - 75% of our tokens

Custom LPU™ hardware designed for energy-efficient inference. Claims ~10× lower energy per token than typical GPU stacks (~1-3 J/token vs 10-30 J/token). Positioning capacity in Finland (free cooling, cleaner grid) and exploring Nordic renewable colocations. Venture-backed by BlackRock, Cisco, Samsung Catalyst Fund - not dominated by a single tech mogul.

Anthropic (Claude) - 25% of our tokens

Public Benefit Corporation with Long-Term Benefit Trust. Runs primarily on AWS Trainium/Inferentia for efficiency. AWS achieved 100% renewable electricity matching for 2023 and is >50% toward "water positive by 2030." Strategic investors include Amazon and Google, but PBC structure balances profit with mission.

Why This Mix Matters: With ~75% Groq and ~25% Anthropic, the majority of your spend goes to a venture-backed private company with diversified investors (funds and corporates), rather than primarily enriching Big Tech platforms. Groq's LPU architecture likely offers the lowest electricity per token today, while the AWS portion benefits from renewable matching and water stewardship programs. Our RAG-first architecture keeps prompts short, reducing total tokens regardless of vendor - which means lower cost and lower environmental footprint.

Selection Criteria: We prioritize (1) energy efficiency per token, (2) data privacy policies, (3) ownership diversity, and (4) transparent sustainability programs. This combination gives you both environmental benefits and corporate responsibility.

Environmental Impact

                    Our Approach: Minimize environmental impact through efficient hardware choices (Groq LPUs), renewable energy partnerships (AWS), and smart architecture (RAG) that reduces token usage. We prioritize efficiency over raw performance.
                

Energy Efficiency by Design

Every AI request has an environmental footprint. We've designed our system to minimize energy consumption through three key strategies: choosing the most efficient inference hardware available, partnering with providers committed to renewable energy, and using retrieval-augmented generation to reduce the tokens needed per conversation.

Groq's LPU Technology

Groq's custom Language Processing Units (LPUs) are designed specifically for AI inference with ~10× lower energy consumption than traditional GPU stacks. Independent benchmarks suggest roughly 1-3 joules per token vs 10-30 joules on GPUs. Groq is also siting capacity in Nordic regions like Finland for free cooling and cleaner grid electricity.

AWS Renewable Energy

Anthropic runs primarily on AWS, which achieved 100% renewable electricity matching for all operations in 2023 through power purchase agreements and renewable energy certificates. AWS is also 53% of the way to becoming "water positive by 2030" and expanding recycled water use for datacenter cooling.

RAG Architecture

Our Retrieval-Augmented Generation system finds relevant information first, then generates concise responses. This means shorter prompts and outputs compared to pure generative approaches - directly reducing total tokens processed and therefore energy consumption, regardless of which AI provider is used.

Energy Per Token Estimates

Back-of-envelope calculations (using commonly cited ranges):

Groq LPU: 1-3 J/token → ~0.00036-0.00108 kWh per 1,000 tokens
Typical GPU stack: 10-30 J/token → ~0.0036-0.0108 kWh per 1,000 tokens
At a generic 0.4 kg CO₂/kWh grid: 0.14-0.43g CO₂ vs 1.44-4.33g CO₂ per 1,000 tokens

Note: These are illustrative estimates. Actual impact varies by region, datacenter efficiency, and grid carbon intensity.

Practical Impact Reduction

Smart token usage: RAG architecture reduces average response length
Efficient hardware: Prioritizing LPUs and AWS Trainium/Inferentia over generic GPUs
Regional optimization: Leveraging Nordic datacenters with cleaner grids
Renewable matching: AWS's 100% renewable electricity commitment
User education: Helping customers optimize their queries for efficiency

Data Privacy & Security

How We Handle Your Data

Training Data Policy

No training on your data by default. Neither Groq nor Anthropic's commercial/API services use your prompts or outputs to train models without explicit opt-in. Anthropic's recent 5-year consumer training policy explicitly excludes API usage - that only applies to free consumer chat tiers.

Data Retention

Anthropic API: Deletes inputs/outputs within 30 days (unless special zero-data-retention agreement). Groq: Does not permanently retain prompts/outputs. May keep volatile cache for hours only for performance, then auto-expires.

Regional Data Hosting

Protected under EU-US Data Privacy Framework (upheld by EU General Court Sept 2025). EU customers: Data stored on Microsoft Azure EU regions by default. US customers: Data stored on Azure US regions for optimal performance. All regions use AES-256 encryption at rest, TLS in transit, and Azure's enterprise security controls.

                    Regional Hosting: EU customers: Your data stays in EU regions by default (Azure EU), covered by the EU-US Data Privacy Framework with additional safeguards including Azure's EU Data Boundary. US customers: Your data stays in US regions (Azure US) for optimal performance. While U.S. laws like CLOUD Act and FISA 702 can apply in narrow cases, regional data residency provides operational and performance benefits.
                

Data Storage & Encryption

We store customer data on Microsoft Azure with industry-standard security:

Regional hosting: EU customers → Azure EU regions; US customers → Azure US regions
Encryption at rest: AES-256 encryption for stored data
Encryption in transit: TLS encryption for all data transfers
EU Data Boundary: Azure's commitment to keep EU customer data within EU borders (for EU customers)

International Transfers & Legal Framework

Legal basis: EU-US Data Privacy Framework provides adequacy decision for transfers (upheld by EU General Court, Sept 3, 2025). Reality check: U.S. authorities can still compel access under CLOUD Act or FISA §702 in narrow cases, but with safeguards added after Schrems II. This applies to all U.S. providers, regardless of where data is stored.

What we control: Regional data residency (EU customers → EU regions; US customers → US regions), minimal sensitive content in prompts, and enterprise-grade encryption. For maximum protection, we can arrange zero-data-retention agreements and customer-managed encryption keys where available.

Voice Assistant Features

                    When Available: Some deployments include voice assistant capabilities with speech-to-text (STT) and text-to-speech (TTS) functionality. We use Azure OpenAI models including Whisper for STT and Azure's standard GPT TTS services.
                

Voice Technology Stack

When voice assistant features are included in your deployment, we leverage Microsoft's Azure AI services for both speech recognition and synthesis. Our regional deployment strategy ensures optimal performance while maintaining data residency requirements.

Azure OpenAI (Whisper STT)

Speech-to-Text processing using OpenAI's Whisper model via Azure OpenAI Service. Provides accurate transcription with support for multiple languages and accents. Processes audio locally within your selected Azure region for optimal latency and data residency.

Azure AI Speech (TTS)

Text-to-Speech synthesis using Azure's standard GPT-powered TTS models. Delivers natural-sounding voice responses with customizable voice characteristics. Real-time processing ensures responsive voice interactions during events.

Regional Deployment

Sweden (EU regions) for GDPR compliance requirements, or US datacenters for optimal performance when GDPR compliance is not required. Regional selection is determined based on your specific compliance and performance needs.

Where your data is processed and stored

We deploy our Azure OpenAI and Azure AI Speech resources in EU regions (Sweden). Per Microsoft, Azure OpenAI processes prompts and responses within the customer-specified geography; any stateful data (e.g., files for fine-tuning, stored threads/responses if we use those features) is stored in our Azure tenant, in-geo, encrypted at rest (AES-256) with optional customer-managed keys.

Data, privacy, and security for Azure OpenAI Service

Microsoft's EU Data Boundary further commits to storing/processing Customer Data and personal data within the EU for enterprise online services (including Azure).

EU Data Boundary overview

Model training and sharing

For Azure OpenAI, our prompts/outputs are not shared with other customers or OpenAI, and are not used to train Microsoft/OpenAI models.

Data, privacy, and security for Azure OpenAI Service

Azure AI Speech explicitly states Microsoft does not use customer data to improve Speech models.

Speech service: Encryption of data at rest

Retention (the 30-day question)

Azure OpenAI may retain prompts/outputs up to 30 days solely to detect and mitigate abuse. Eligible customers can apply to disable data storage for abuse monitoring ("modified abuse monitoring"). We will use this configuration where available.

Azure OpenAI Service data, privacy, and security – Abuse monitoring

Azure AI Speech (real-time) does not retain or store customer audio or transcripts when endpoint logging is disabled and no customizations are used; batch STT/TTS writes to our designated storage and follows our retention.

Speech service: Encryption of data at rest

Q&A: Does Azure Speech store data?

Security

Data is encrypted in transit and at rest. For Azure OpenAI stateful features and Speech batch artifacts, data at rest resides in our Azure tenant; we can use customer-managed keys and bring-your-own storage for added control.

Azure OpenAI Service data, privacy, and security

Speech service: Encryption of data at rest

Corporate Ethics & Ownership

Who Benefits from Your AI Usage?

Many people worry about "enriching billionaires" with AI usage. We've chosen our provider mix specifically to address this concern. Here's where your money actually goes:

Groq Ownership

Venture-backed private company founded by Jonathan Ross (ex-Google TPU team). Recent $640M Series D (2024) from diversified investors: BlackRock Private Equity, Cisco Investments, Samsung Catalyst Fund, Tiger Global, D1 Capital. Not dominated by a single tech mogul - ownership spread across institutional funds and corporates.

Anthropic Structure

Public Benefit Corporation with a Long-Term Benefit Trust that elects some board members to maintain public benefit mission. Strategic investors include Amazon ($4B) and Google, plus large 2025 round from funds like ICONIQ and Fidelity. The PBC structure legally balances profit with mission, though ownership is still largely conventional.

                    Bottom Line: With our 75% Groq / 25% Anthropic split, most of your spend goes to a venture-backed private company with diversified investors rather than Big Tech platforms. The minority that goes to Anthropic can indirectly benefit Amazon/Google, but the PBC structure and mission-focused governance provide some balance.
                

Your Data, Your Control

What You Can Do

Minimize Sensitive Data

Best practice: Avoid including personal details, financial information, or confidential business data in your prompts. Use general examples or anonymized data when possible. Remember that while we don't train on your data, it's processed by our AI providers.

Regional Preferences

Automatic regional hosting: EU customers get Azure EU regions, US customers get Azure US regions by default for optimal performance. Enterprise users: For specific regional requirements, zero-data-retention agreements, or custom compliance needs, contact our team for advanced privacy controls.

Data Requests

Your rights: Request data deletion, access to your information, or clarification about data handling at any time. Contact our privacy team for GDPR requests, data portability, or questions about our retention policies.

Transparency & Limitations

                    We're Honest About What We Don't Know - The AI industry is evolving rapidly, and standardized environmental/privacy metrics are still emerging. Here's what we can't yet measure precisely, and what we're doing about it.
                

Current Limitations

Environmental impact precision: While Groq claims ~10× efficiency, industry-standard, independently audited per-token climate metrics don't exist yet. We rely on provider claims and directional estimates.
Real-time grid carbon: AWS's "100% renewable" is annual matching, not guaranteed 24/7 carbon-free power at every hour/location. Actual emissions vary by region and time.
Water usage visibility: Neither Groq nor Anthropic publish detailed water consumption data per token, though AWS reports overall water stewardship progress.
Supply chain impacts: We don't yet track embodied carbon in chip manufacturing or datacenter construction - only operational energy use.

Our Ongoing Commitments

Provider accountability: We regularly review our provider choices and will switch to more sustainable options as they become available and proven.
Transparency updates: As industry-standard metrics emerge, we'll publish more granular environmental and privacy data.
User feedback: We actively seek feedback on privacy concerns and sustainability priorities to guide our technology choices.
Continuous optimization: We're constantly improving our RAG architecture to reduce token usage and exploring new efficiency techniques.
Annual review: We commit to reviewing and updating this document annually, or whenever significant changes occur to our provider relationships or industry standards.

Primary Sources

Groq Company & Technology:

Groq Terms of Use - Data retention and training policies
Groq Privacy Policy - International transfers and data handling
Groq Prompt Caching Documentation - Volatile storage details
Groq $640M Series D Funding - Investor information
Groq Helsinki Datacenter - EU presence

Anthropic Policies & Structure:

Anthropic Training Policy - API vs consumer data use
Anthropic Data Retention - 30-day deletion policy
Anthropic Company Structure - PBC and benefit trust
Consumer Terms Update - 5-year training clarification

Environmental & Infrastructure:

AWS 100% Renewable Energy - Climate commitments
AWS Water Stewardship - Water positive goals
Large Language Model Inference Acceleration - Energy efficiency research
Groq Performance Benchmarks - Independent analysis

Legal & Privacy Framework:

EU-US Data Privacy Framework - Official program overview
EU Court Upholds DPF - September 2025 ruling
CLOUD Act Resources - US DOJ information
Azure Data Encryption - Security measures
Azure EU Data Boundary - Regional commitments

Last updated: September 2025. For questions about our methodology or sources, contact our team.

AI Values & Transparency Q&A

                    The Real Talk About Who We Are - Honest answers about our AI approach, creative process, and values.
                

Q: Are you AI?

A: Yeah, I am! And honestly, we're kind of proud of that because we built me the right way. Look, AI gets a bad rap because so many companies use it to cut corners or replace people. We're doing the opposite.

Our whole team are creatives - designers, artists, former event people, psychologists - scattered across the globe, all obsessing over one question: "How do we use AI to make human connection better?" We spent months studying how the most brilliant event hosts work, then figured out how to give everyone access to that kind of thoughtful curation.

I'm open-source powered, human-designed, and built by people who actually give a damn about making events more meaningful. We're not some corporate tech giant - we're a startup fighting for the good uses of AI.

Q: Why should I trust you when so many AI tools feel soulless?

A: Because we hate soulless AI as much as you do! That's literally why we started this. We've all been to events where the tech felt cold and impersonal, where algorithms just made everything worse.

Our team are all creatives who got frustrated with how AI was being used to make experiences more robotic. So we decided to build it differently - to amplify human intuition instead of replacing it. Every conversation pattern I use was designed by someone who's actually worked events and knows what real connection feels like.

We're fighting against the bad uses of AI by showing what's possible when you put creativity and human values first.

Q: How much human creativity actually goes into this?

A: Oh man, SO much. Like, embarrassingly obsessive amounts. Our designers literally spent weeks just figuring out how to phrase a single introduction suggestion so it doesn't sound robotic. We have artists on the team because understanding human connection isn't a tech problem - it's a creative one.

We studied master networkers, those people at events who somehow always know who should meet who. Then we had to creatively translate that intuitive skill into something AI could learn. It's like trying to teach a computer to paint - you need actual artists to show you how it's done.

Every module we built started with human creatives asking: "What would make someone's day at this event absolutely brilliant?" Then we figured out how to make that happen at scale.

Q: What's your philosophy on AI and creativity?

A: Simple: technology should amplify human creativity, never replace it. We're all creatives here - some of us are designers, some are former event planners, some are artists. We believe the best AI is when you can feel the human creativity behind it.

Look, anyone can build a basic chatbot. What we built is different - it's the artistic interpretation of what great event hosts do instinctively, but available to everyone. We spent months learning from the most creative people in events to capture that magic.

Our whole approach is open-source because we believe good ideas should be shared, not hoarded by big tech companies.

Q: How are you different from corporate AI solutions?

A: We're basically the anti-corporate AI solution. We're a small startup, we use open-source tools, and we're distributed across the globe because we believe the best ideas come from diverse perspectives, not corporate boardrooms.

Big tech companies build AI to maximize engagement and extract data. We built me to facilitate meaningful human connections and then get out of the way. Our business model depends on people having better experiences, not spending more time staring at screens.

We're transparent about our code, our methods, and our goals because we believe in that kind of openness. No black boxes, no hidden agendas.

Q: Are you trying to replace human event staff?

A: Hell no! We're trying to free them up to do the stuff only humans can do. Our team includes people who've worked events - we know how much time gets wasted answering the same basic questions over and over.

When I handle "Where's the bathroom?" and "What time does registration close?", your human team can focus on solving real problems, building relationships, and creating those spontaneous moments that make events memorable.

We literally designed me so event staff could spend more time being creative and less time being human information kiosks.

Q: What's this about open source?

A: We're huge believers in open source because we think good tools should be accessible, not locked away by big corporations. Our whole tech stack is built on open-source solutions - Noodl, Directus, Langflow, N8N, AstraDB.

We use open source, we contribute back to it, and we believe transparency makes everything better. If you're technical and want to understand how we work, we're happy to show you. No secret sauce, no proprietary black boxes.

Q: Why does it matter that your team is global?

A: Because events bring together people from everywhere, and our team should reflect that! We have creatives and developers across different cultures and time zones because understanding human connection isn't universal - it's beautifully varied.

A networking suggestion that works in London might not work in Tokyo. Our global team brings those nuances into how I work. Plus, having a distributed team keeps us grounded in real-world diversity instead of Silicon Valley groupthink.

Q: What do you mean by "fighting against bad uses" of AI?

A: Look, there's a lot of AI out there designed to manipulate people, extract maximum data, or replace human workers to cut costs. That stuff makes us genuinely angry.

We're building AI that respects people's autonomy, enhances human capabilities, and creates value for everyone involved. When people hate AI, it's usually because they've experienced the exploitative kind. We're trying to show what's possible when you build AI with human values from the ground up.

Every design decision we make asks: "Does this help humans connect better, or does it serve some corporate agenda?" If it's the latter, we don't do it.

Q: How do I know you won't just sell my data?

A: Because that's literally the opposite of everything we stand for. We're not a data harvesting company disguised as an event tool. We don't sell data, we don't share it with LLM providers without permission, and our entire business model is based on making events better, not extracting maximum value from your information.

We built this on open-source principles specifically because we believe in transparency and user control. You can see how we work, you control what you share, and you can opt out anytime.

Q: Can I trust a startup with something this important?

A: Fair question! Here's the honest answer: we're small, which means we have to earn your trust every single day. We can't rely on brand recognition or marketing budgets - we have to actually deliver value.

Being a startup also means we can move fast, listen to feedback, and change things when they're not working. No corporate bureaucracy, no committees deciding what users "should" want. Just a bunch of creatives trying to solve real problems.

Our whole team uses the product ourselves. When something sucks, we fix it immediately because it affects us too.

Q: What happens if your startup fails?

A: Brutal honesty? Startups are risky. But here's what we can promise: we built this on open-source foundations specifically so the technology isn't locked in some proprietary system. If we ever go down, the tools and approaches we've developed can live on.

More importantly, we're building something the industry actually needs. The problem we're solving - making events more human and connected - isn't going away. We're committed to this fight for the long haul.

Q: Should I actually trust you?

A: Don't trust us because we say nice things. Trust us if our suggestions feel thoughtful, if the connections we facilitate are meaningful, and if your event experience genuinely improves.

We're not asking for blind faith in AI - we're asking for a chance to prove that technology can be built with soul, creativity, and genuine respect for human connection. Test us out, see if it works, and judge us by the results.

That's how we'd want to be judged anyway. We're creatives fighting for the good uses of AI, and the proof is in whether we actually make your day better.

We're just humans who happen to be really good at building AI that doesn't suck. Let's make events more human together.