AI security and control concerns dominate VivaTech in Paris

Europe’s largest startup and technology conference put AI’s practical risks ahead of its flashiest promises this year. Fortune reported that VivaTech in Paris drew 180,000 attendees and centered on cybersecurity, national control of AI infrastructure and pressure from companies to prove returns on AI spending.

According to Fortune’s Beatrice Nolan, the event still had its futurist moments. Jeff Bezos argued that AI would create a labor shortage rather than broad job losses and described a longer-term goal of moving heavy industry off Earth and building an industrial presence on the Moon.

Yann LeCun, executive chairman of AMI Labs and former head of Meta’s AI work, offered a more cautious view. Fortune cited LeCun’s CNBC comments that the AI industry could face a “big bubble explosion” if companies cannot reduce costs quickly enough.

Fortune reported that the conference’s more memorable counterpoint came from two humanoid robots whose booth demonstration went wrong when they backed into televisions and knocked two screens to the floor. Away from the exhibition floor, Nolan wrote, executives focused on nearer-term problems.

Cyber risks move to the front

Fortune reported that Anthropic’s Mythos and OpenAI’s GPT-5.5-Cyber have increased concern among business and government leaders because of their ability to assist with advanced hacking work. Nolan wrote that such models can identify unknown software flaws and produce working exploits at a speed and scale that previously required specialized human effort.

According to Fortune, Anthropic and OpenAI have tried to slow broad access to some cyber-capable products by first giving vetted security firms and critical infrastructure operators access. The goal, Fortune reported, is to let defenders find and fix vulnerabilities before those capabilities spread more widely.

Peter DeSantis, a senior vice president at Amazon, told Fortune that the long-term balance may favor defenders, while the near-term risk remains a concern. He said security teams are still trying to understand the attacks these systems make possible and update their tools and practices.

Fortune also reported that OpenAI expanded its Daybreak initiative with Patch the Planet, an open-source patching effort tied to GPT-5.5-Cyber. The effort involves firms including Cloudflare, Cisco and CrowdStrike and is aimed at helping organizations detect and repair vulnerabilities.

Sovereignty debate sharpens

Fortune reported that a U.S. decision to cut off access to Anthropic frontier models gave new urgency to Europe’s discussion of “sovereign AI.” Nolan wrote that the term meant different things to different executives and policymakers at VivaTech.

Cohere CEO Aidan Gomez told Fortune that sovereignty should begin with domestically controlled infrastructure, including chips, power, data centers and private deployment. He said countries lacking that capacity should form strategic alliances to counter U.S. and Chinese control over key AI infrastructure.

DeSantis gave Fortune a less rigid view, saying no country, including the United States, has fully sovereign infrastructure. He argued that countries should focus on keeping sensitive data inside national borders and maintaining clear governance over how AI is used, rather than trying to rebuild every layer of the supply chain.

Companies ask for results

Fortune reported that many corporate discussions at VivaTech turned to training, automation and the cost of AI programs. Philippe Rambach, Schneider Electric’s chief AI officer, told Fortune he had made AI training mandatory for the company’s 42,000 employees.

Rambach said Schneider Electric was being selective about AI pilots and backing projects with defined business cases and a route to broader deployment, according to Fortune. Nolan wrote that the company is treating AI more as an operating tool than as a showcase technology.

OpenAI is seeing similar questions from enterprise customers, according to Thibault Sottiaux, the company’s head of core platforms. He told Fortune that clients often ask whether AI agents are producing enough value for the money spent, and said OpenAI is focused on making models more efficient so companies can do more with lower costs.

This story draws on original reporting from Fortune.