AI is speeding up cyberattacks but core techniques remain phishing, malware and exploits
- Severity
- Medium
- Exploited
- Yes
- Sources
- 1
What happened
Experts argue AI is compressing attacker timelines and lowering the cost of operations, while core tactics such as phishing, malware and exploitation persist. JadePuffer, a ransomware campaign disclosed in July 2026, used an LLM-driven agent and initially exploited CVE-2025-3248, a year-old missing-authentication flaw in internet-facing Langflow instances, to harvest cloud credentials and API keys before destructive database encryption. Defenders should prioritise exposure checks and patching of AI and agent frameworks, and assume shorter dwell times, since detection windows are shrinking.
For security analysts · Written by AI from 1 source · Wrong?
What to do
- Check whether any internet-facing Langflow instances are in use, as CVE-2025-3248 was used for initial access.
- Prioritise patching of exploitable software flaws, particularly in open-source AI frameworks.
- Review credential and API key exposure in systems that run AI workflows.
Written by AI from the sources, checked against them
Vulnerabilities
Full story timeline · 2 updates
Who reported · 1 outlet
Who and what is in it
Products