🧔♂️ A friendly human may check it before it goes live. More news here
Israel’s cyberattacks surge as AI tools add new risks: report
Israeli organizations face an average of 2,000 cyberattacks per week, according to new data from Check Point, a cybersecurity firm based in Tel Aviv.
Telecommunications in Israel saw the highest activity, with about 4,000 attacks per week per organization, compared to a global average of 2,703.
The transportation and logistics sector reported 3,017 weekly attacks per organization, more than double the global figure of 1,169.
Check Point’s CPO, Nataly Kremer, shared these findings during Cyber Week at Tel Aviv University.
Kremer noted that rapid adoption of AI tools has created new security risks, citing the discovery of around 10 major flaws in AI applications in a single month.
She highlighted vulnerabilities such as training data manipulation, information leaks, and hidden weaknesses in AI development environments.
Kremer emphasized the need for dedicated security measures as organizations integrate AI into their operations.
🔗 Source: Calcalist
🧠 Food for thought
Implications, context, and why it matters.
Israel’s attack volume lacks context without trend data or methodology
- That 2,000 attacks per organization each week is hard to assess without Check Point methods, sample size, or time-series data to see if this rose or fell 1.
- Telecom logs 4,000 attacks a week against a 2,703 global average, while transport and logistics see 3,017 vs 1,169 1. The meaning of an incident and per-organization math across sizes is unclear. We also lack data on severity, success, or bot probes versus breaches.
- Check Point sells only through partners (selling primarily through resellers/partners rather than directly to end customers) and Israel hosts 500+ cyber firms (a large domestic cybersecurity industry), which can skew threat reports 12.
Security auditors can offer Large Language Model (LLM) red teaming services using the Open Worldwide Application Security Project (OWASP) frameworks
- Ten AI application flaws found in one month match OWASP guidance on LLM risks 34. It covers prompt injection (malicious inputs that steer model behavior), training data poisoning (tampering with data used to train models), and system prompt leakage (exposing hidden instructions).
- Cyber firms can sell AI red teaming services (controlled adversarial testing) for OWASP Top 10 risks 4. Emphasize prompt injection and sensitive information disclosure, which tend to be easy to exploit with high impact.
- Consultants can build Retrieval-Augmented Generation (RAG) security audits that probe vector and embedding weaknesses (vectors and embeddings are numerical representations of text used for retrieval) 45. These map to a 2025 OWASP category on unauthorized access and embedding inversion attacks (reconstructing sensitive text from embeddings) in retrieval systems.
- Service providers can offer excessive agency testing for agentic AI systems (models that can autonomously take actions via tools and APIs) 65. Align this with the OWASP Top 10 for Agentic Applications 2026 framework (a risk list created with 100+ industry experts).
Stay updated on the go with our mobile app.
Get latest insights with smoother, more personalized experience through TIA mobile app.