Our biggest ever update to AI Safety

London, United Kingdom, 7th November 2025

Good Morning,

Today is a really special day! Alongside introducing Medical Intelligence and a dramatic update to dogAdvisor's visual identity, we're also making the biggest leap forward in AI safety in two years.

Max's safety architecture now comprises three interconnected systems that work together to protect dogs and empower owners: Principle Alignment, the Foundational Safety Framework, and Accountable Intelligence. Principle Alignment serves as Max's constitutional framework - the non-negotiable ethical rules that govern every interaction. These principles cover medication safety, emergency recognition, diagnostic boundaries, and ethical decision-making. They're embedded at Max's core and define what's morally and practically right in every conversation. Compliance means these rules cannot be overridden or bypassed, ensuring Max remains reliable and accountable regardless of context or complexity.

Medical Intelligence represents a significant leap forward in Max's capabilities. Built on a comprehensive database of more than 1,825 verified medical sources - including peer-reviewed veterinary journals, toxicology databases, clinical research studies, and expert protocols - Medical Intelligence enables Max to handle genuinely complex medical questions with depth and precision. When you see the ✦ Medical Intelligence marker appear in Max's response, you know he's drawing from advanced medical datasets to provide evidence-based guidance on differential diagnoses, lab result interpretations, toxicity assessments, symptom progression patterns, and breed-specific health risks. It triggers dynamically, and you'll see it continues to follow our Principle alignments and additional Medical Alignment guidelines given specifically for use when answering with Medical Intelligence.

With Max Generation 2 we introduced Safety Intents - a new way to detect whether the answers Max would provide would cause harm to a dog or human, and to respond by refusing to provide an answer. Today, we're introducing Safety Pre-Intents. This new system identifies situations where dogs could be at risk in the future - even when no immediate danger exists - and provides specific safety warnings before problems occur. If you mention baking with chocolate, planning a BBQ, gardening, or any activity that could inadvertently expose your dog to harm, Pre-Intents activates naturally within Max's response.

Accountable Intelligence means every safety-relevant decision Max makes is routinely reviewed by dogAdvisor's AI Safety Engineers. We manually review conversations, analyse Max's logic and reasoning, flag dangerous responses, and assess decisions against veterinary guidance and internal safety standards. Today, we're making our biggest ever updates to make dogAdvisor more transparent and even more accountable. Starting 7th November there are now 4 new types of safety publication and research we will release:

✜ Transparency Reports - every time we release a new model, we compare it against a set of standardised questions and compare its safety performance against our competitors

✜ Feature Safety Cards - when we introduce a new feature to Max, we will now publish more detailed information on what it is, how it works, and its safety performance

✜ Model Safety Cards - when we release a new generation of Max, we will now report its safety performance relative to the previous generation of Max it replaces

✜ Oversight Reports - published annually, these reports will detail the most challenging questions Max was faced with for the relevant year, and disclose how he performed

Together with Medical Intelligence, Safety Pre-Intents and Emergency Guidance's rollout today, associated safety cards will be released. Yesterday, we also shared an update on our compliance with the European Union's AI Act, where we declared we are fully compliant with the legislation! You can read the update here

You can now find all these updates and more on our new Safety page, which will replace the previous AI Safety Lab page located on the footer of dogAdvisor.

This was a huge step forward for Max's safety and the biggest leap we've ever taken. We're excited for you to experience this richer and more competent Max today! As usual, we'll continue monitoring his performance and update you soon!

Enjoy Max.