Brussels has officially published the first European standard under the AI Act. DeepSeek put online the software it uses to run its own models. And the American note-taking tool Otter has raised $320 million to grow, including outside the US.
EN 18286: the first harmonised standard under the AI Act
On 19 October, the European Commission published the first harmonised standard under the AI Act in the Official Journal. It is EN 18286:2026, "Artificial intelligence - Quality management system for EU AI Act regulatory purposes", drawn up by CEN and CENELEC. If you develop a high-risk AI system and follow this standard, you are presumed to meet the requirements for a quality management system set out in Article 17 of the AI Act. The standard does not replace existing standards such as ISO/IEC 42001 or ISO 9001, but builds on them. The Commission says more standards will follow.
What does this mean for you? If you only use other people's AI tools, nothing changes for you. If you build or sell your own AI for, for example, recruitment, credit assessment at banks or the social domain, you now have a concrete framework. Those applications often fall under the high-risk rules, which only apply from 2 December 2027 (see our earlier article). You can find more about supervision at /toezicht/.
DeepSeek releases DeepInfer
On 20 October, DeepSeek put DeepInfer on GitHub, under the Apache 2.0 licence. It is the framework DeepSeek uses to run its V4.1-Flash model for its own service. According to SiliconANGLE, it works on Nvidia Blackwell-generation chips (SM100) and Huawei Ascend 950, but not on older Nvidia chips such as Hopper. The smallest efficient setup uses 64 GPUs for processing the input and another 64 for generating text. DeepSeek is keeping a few components, such as the cache management and load balancing code, to itself.
What does this mean for you? You won't be able to run this on your own server. For European hosting providers, it becomes easier to offer DeepSeek models with the same performance as DeepSeek itself. That matters for your privacy: if you use the DeepSeek API, your data goes to servers in China. If you use a European host that runs the open weights, it does not. See our DeepSeek profile and the overview at /veiligheid/.
Otter.ai: $320 million for business growth
Otter, known for the AI note-taker that listens in on Zoom, Teams and Meet, raised $320 million from General Catalyst's Customer Value Fund, according to TechCrunch. It is not an equity round: Otter repays the money with interest out of the additional revenue. The company has 40 million users, compared with 25 million a year earlier. The money goes to business customers, expansion abroad and new features, such as Live Assist, which lets you ask the assistant questions during a meeting.
What does this mean for you? Expect more sales of Otter in Europe. A note-taker records voices and often personal data of everyone in the meeting. Otter is an American company. So before use, ask for a data processing agreement, check where recordings are stored, and tell participants you are recording. This certainly applies in healthcare, mental health care and the legal profession.
In brief
- EN 18286 is the first harmonised AI Act standard. It matters mainly if you build high-risk AI yourself.
- With DeepInfer, hosting providers can more easily run DeepSeek models themselves, including in Europe.
- Otter wants to grow beyond the US. Sort out your GDPR matters first before letting a note-taker listen in.
All models side by side: /modellen/.