/
← Voltar ao feed
medium

Trend Micro survey: 94% of firms expect AI to raise cyber risk; Pwn2Own Berlin AI category yields 7 zero-days

1 fonte(s) · 1 atualização(ões) · publicado 01/07/2025 15:55 · coletado há 1 dia

Trend Micro published research, conducted by Sapio Research among 2,250 IT and security decision-makers in 21 countries, on AI adoption in cybersecurity. 81% of businesses already use AI-driven tools in their security strategy, but 94% expect AI to negatively affect their cyber risk exposure within three to five years. Over half expect a surge in the scale and complexity of AI-driven attacks. Respondents cited sensitive data exposure, uncertainty over how AI systems process and store data, proprietary data exploited by untrusted models, compliance pressure, and monitoring challenges from new endpoints, APIs and shadow IT.

Technical Details

  • The first AI category at Pwn2Own Berlin drew twelve entries targeting four AI frameworks, with NVIDIA Triton Inference Server attracting the most attention.
  • Chroma, Redis and the NVIDIA Container Toolkit were also successfully exploited, in some cases with a single bug achieving full compromise.
  • Seven unique zero-day vulnerabilities were found. No CVE identifiers or technical details are given in the article.
  • Vendors have 90 days to patch before technical details are published.

Remediation

  • Trend urges security leaders to evaluate the evolving AI risk landscape and build security into every stage of AI adoption.
  • Apply vendor patches for the affected AI frameworks once they are released.

Linha do tempo

  1. newsroom.trendmicro.com ↗ · pub. 01/07/2025 15:55 · coletado 19/10 18:02 primeiro registro

    Trend Micro survey: 94% of firms expect AI to raise cyber risk; Pwn2Own Berlin AI category yields 7 zero-days

    ler a análise desta fonte

    Trend Micro published research, conducted by Sapio Research among 2,250 IT and security decision-makers in 21 countries, on AI adoption in cybersecurity. 81% of businesses already use AI-driven tools in their security strategy, but 94% expect AI to negatively affect their cyber risk exposure within three to five years. Over half expect a surge in the scale and complexity of AI-driven attacks. Respondents cited sensitive data exposure, uncertainty over how AI systems process and store data, proprietary data exploited by untrusted models, compliance pressure, and monitoring challenges from new endpoints, APIs and shadow IT.

    Technical Details

    • The first AI category at Pwn2Own Berlin drew twelve entries targeting four AI frameworks, with NVIDIA Triton Inference Server attracting the most attention.
    • Chroma, Redis and the NVIDIA Container Toolkit were also successfully exploited, in some cases with a single bug achieving full compromise.
    • Seven unique zero-day vulnerabilities were found. No CVE identifiers or technical details are given in the article.
    • Vendors have 90 days to patch before technical details are published.

    Remediation

    • Trend urges security leaders to evaluate the evolving AI risk landscape and build security into every stage of AI adoption.
    • Apply vendor patches for the affected AI frameworks once they are released.