Google backs 33 cybersecurity startups in AI shift

Google has selected 33 cybersecurity startups for Gemini Startup Forum: Cybersecurity, the programme's first cohort.

The forum brings together companies working across six areas of cyber defence, with a strong focus on artificial intelligence in security products and operations. Participants will work with specialists from Google DeepMind, Google Cloud and Wiz as they develop their products.

The list highlights how quickly security vendors are shifting towards tools built for autonomous AI systems, as well as conventional corporate networks, cloud environments and endpoints. Several of the selected companies focus on monitoring, governing or restricting the actions of AI agents, a segment that has emerged as businesses test software that can act with a degree of autonomy.

Among the startups in AI agent security and governance are Capsule Security, Evoke Security, Manifold Security, Mirror Security, Onyx Security, Refractal, Unbound Security and XOR. Their products range from runtime monitoring and behavioural analysis for AI agents to encrypted processing, audit logging and training data for secure coding systems.

That concentration reflects a broader shift in cybersecurity priorities. As organisations deploy generative AI systems and agent-based software in business processes, they create new attack surfaces tied to identity, data access, model behaviour and software permissions.

Six areas

The selected startups span application security and vulnerability management; cloud, network and infrastructure security; endpoint security and data protection; security operations centre automation and offensive security; security infrastructure and specialist services; and AI agent security and governance.

Application security companies in the group include Aisy, Alt Security, Arcjet and Pixee. Their work covers attack-surface modelling, automated penetration testing, in-code application protection and the automatic conversion of vulnerability scanner findings into code fixes.

Cloud and infrastructure security is another large segment in the cohort. In that category, Google named CloudFence, CyberSeQ, Huskeys, Native, Prowler, QIZ Security and Tracebit. Several target multicloud policy management, misconfiguration detection and preparation for post-quantum cryptography, showing how infrastructure defence is expanding beyond traditional monitoring tools.

On the endpoint and data protection side, Bold Security, Glow, Jazz, ORION Security and MokN were selected. Their products focus on data loss prevention, insider risk, endpoint activity monitoring and credential theft.

The security operations centre automation and offensive security category includes COGNNA, Latent Defence, Mate Security, Nrdsnipe, RIFFSEC and TandemTrace. Many use AI to sift alert volumes, model attack paths or automate parts of incident investigation and internal penetration testing.

In security infrastructure and specialised services, the selected companies are Netsec, Revelum and Synqly. Their work spans IT and security administration, detection of deepfake-led fraud and software integration between cyber tools.

Regional spread

The cohort is geographically diverse, with startups from the United States, Israel, the United Kingdom, Ireland, France, Saudi Arabia, Sweden, Poland and Spain. The United States and Israel account for a large share of the list, reflecting their long-standing role in cybersecurity startup formation. The UK also features strongly across AI governance, vulnerability management, infrastructure protection and deception-based detection.

The selection also points to several technical themes drawing investment and customer attention. One is post-quantum cryptography, with CyberSeQ and QIZ Security both building tools to help organisations identify and manage cryptographic assets as they prepare for new encryption standards. Another is deception technology, visible in products from Tracebit and MokN, which use decoys to identify intrusions or expose stolen credentials.

There is also a notable push towards automation in security operations. Startups such as COGNNA, Mate Security and TandemTrace aim to reduce the burden on analysts by using AI agents to investigate alerts and connect disparate signals from existing security platforms. That approach has gained traction as security teams struggle with staff shortages, high alert volumes and increasingly complex IT estates.

Over the past four years, Google says it has backed more than 50 cybersecurity founders through its startup efforts, including companies such as Authologic, BforeAI, Build38, Cerby, Crowdsec and Risk Ledger. The latest cohort suggests that support is moving more directly towards companies building controls around autonomous software, alongside those focused on cloud posture, data protection and security operations.

The make-up of the forum also shows how cybersecurity product design is changing. Rather than focusing only on perimeter defence, many newer companies are building around governance, visibility and response within distributed systems where users, machines, applications and AI agents all interact.

Google says the cohort reflects the industry's shift from perimeter defence to autonomous AI agent security, with an emphasis on enterprise governance and the safe deployment of self-governing AI.