RECIF 2026: AI and the new challenges of Industrial Cybersecurity
October 8, 2026 · News
Key takeaways
- On September 23-24, Wavestone attended RECIF 2026, the event dedicated to industrial cybersecurity, held in Évian-les-Bains. This year’s edition, themed “Accelerating or Slowing Down Industrial Digitalization: Choose Your Side” brought together cybersecurity experts and industrial stakeholders to discuss the challenges associated with the digital transformation of industrial environments.
- During the event, Gérôme Billois and Alexandrine Torrents took part in the panel discussion “Industrial Digital Governance: Is There a Pilot in the Cockpit?”, which focused on governance challenges related to the digital transformation of industry.
- Wavestone also participated in the RECIF Speed Demos with a new demonstration of Picky Factory, our industrial mock-up designed by Alexandrine Torrents and Arnaud Soullié.
When conversational AI meets the industrial world
Picky Factory showcases a waste-sorting plant modernized through artificial intelligence. Combining waste visual recognition, robotics, edge computing, MQTT, AWS cloud services, and a conversational AI agent, the model replicates a modern industrial architecture in which a technician can interact with the production line using natural language.
However, integrating AI into industrial environments also introduces new risks.
In the scenario presented at RECIF, hacktivists discover an online demonstration of the system. By exploiting a prompt injection vulnerability, they are able to bypass the AI agent’s safeguards and interact directly with the production system.
They then discover that one of the agent’s functionalities has write permissions on the system responsible for transmitting sorting rules to the production line. With a simple natural-language instruction, they are able to modify these rules and directly affect the industrial process.
Securing AI by Design
Beyond the scenario itself, Picky Factory helps raise awareness among IT, OT, and cybersecurity teams about the new challenges associated with integrating AI into industrial environments. The demonstration highlights two key principles: separating internet-exposed demonstration environments from production environments, and limiting AI agent permissions to the strict minimum required.
As AI adoption continues to accelerate across industry, security must be incorporated from the earliest stages of designing architectures that combine AI, cloud, edge computing, and industrial systems. AI Cyber by Design is becoming a critical requirement for supporting the transformation of industrial environments while maintaining control over the new risks that come with it.
