HardenStance Says Deutsche Telekom Led Telco Peers on AI Security at MWC 2026

Summary

HardenStance’s MWC 2026 briefing praises Deutsche Telekom’s AI security vision, finds Telefónica’s efforts less prominent, and says Orange showed little evidence of AI risk mitigation at its booth.

Key points

HardenStance describes Deutsche Telekom’s five-stage AI security vision, spanning governance, AI defense and identity.
Deutsche Telekom said AI agents need assigned identities and least-privilege permissions, and that MCP and A2A protocols require security.
The company described using AI agents for chatbot testing and monitoring natural-language attacks, and said it had deployed 6,000 honeypots to collect threat data on LLM targeting.
Deutsche Telekom planned to test its Mobile.ID secure chip with employees in spring 2026, with a commercial launch targeted for summer and initial support for Samsung smartphones.
Deutsche Telekom and Google Cloud demonstrated MINDR, a multi-agent system for autonomous diagnostics and operations across telecom networks.
The briefing says Telefónica presented AI security and identity themes but offered less connection to an AI-agent-focused roadmap, while Orange’s booth showed little evidence of AI risk mitigation.

Article Details

Publisher: HardenStance
Report Period: March 2026
Scope: Assessment of telco security and AI-security positioning at MWC 2026, including Deutsche Telekom, Telefonica, and Orange.
Key Statistics: Deutsche Telekom said it had deployed 6,000 honeypots throughout its systems to collect threat data on attacks targeting LLMs.
Deutsche Telekom said AI agents enable it to test chatbots 20 times faster.
Of Orange's 11 MWC demos, the report says three addressed an aspect of risk mitigation.
Deutsche Telekom said its platform manages the identities of 200,000 employees.
Deutsche Telekom presented a five-stage AI-security roadmap.
Recommendations: Secure agentic AI, including the Model Context Protocol and Agent-to-Agent protocols, as rigorously as other software.
Red-team and test LLMs, including using AI agents to interrogate chatbots.
Monitor natural-language content for attacks that seek to exfiltrate sensitive information.
Assign AI agents identities and unique permissions, applying least-privilege access.
Use cryptography for verification and authentication, including embedding cryptographic keys in AI agents.