ReversingLabs reports AI-linked malware signals rising, no clear attack surge
ReversingLabs says LLM API references in malicious files are rising but no AI attack surge is shown; review exposure to PHP2Shell CVEs.
- Published
- Exploitation
- Not confirmed. ReversingLabs says Amazon reported an attempted exploit against its honeypots before disclosure, and that some circulating proof-of-concept code did not achieve full code execution.
Live scores
CVE-2026-109618- EPSS
- 1.000
very likely to be attacked soon - VEDAS
- 0.982
strong exploit evidence - On CISA KEV
- Yes, since 2026-10-26
Thin horizontal line: the “likely exploited” threshold (EPSS 0.10, VEDAS 0.5). Shaded “no data” areas are periods when cve-scores did not publish. The vertical line is the day CISA added it to its exploited list.
VEDAS flagged this bug about 5 days before CISA confirmed attacks.
Live scores
CVE-2026-109619- EPSS
- 1.000
very likely to be attacked soon - VEDAS
- 0.987
strong exploit evidence - On CISA KEV
- Yes, since 2026-10-26
Thin horizontal line: the “likely exploited” threshold (EPSS 0.10, VEDAS 0.5). Shaded “no data” areas are periods when cve-scores did not publish. The vertical line is the day CISA added it to its exploited list.
VEDAS flagged this bug about 2 days before CISA confirmed attacks.
What happened
According to a ReversingLabs blog post by Igor Lasic, SVP of Engineering, the predicted wave of autonomous, AI-driven cyberattacks has not yet shown up as a clearly measurable surge in cyber losses, but warning signs are growing ReversingLabs. This is one vendor’s analysis, and no independent verification of its figures is given in the source.
ReversingLabs says that in an October 24 search of its malicious-file corpus, an estimated high-confidence subset of files containing references to large language model (LLM) APIs grew from roughly 150 in 2024 to 620 in 2025 and 1,680 from January 1 through October 24, 2026. The company says the broader search returned 596 malicious matches in 2024, 3,044 in 2025 and 10,821 through October 24, 2026. It states that these are early signals, not a count of AI-orchestrated intrusions, and that an API reference does not prove model execution, attacker intent or successful exploitation. It also says a jump in monthly matches from May to June 2026 has not been tied to a specific malware family, collection change or detection change.
ReversingLabs reports samples of three AI-enabled malware families described by other researchers: PromptFlux (11 samples), LameHug / LameHubloader (23 samples) and PromptSpy (3 samples). It says actor attributions come from the cited outside reports and that it has not independently attributed the samples.
On PHP2Shell, ReversingLabs says the two vulnerabilities, CVE-2026-109618 and CVE-2026-109619, were the subject of AI-assisted discovery and rediscovery reporting. It says Amazon reported an attempted exploit against its honeypots before public disclosure, while other researchers cautioned that some circulating proof-of-concept code did not achieve full code execution. ReversingLabs says it identified 10 PHP2Shell exploit files, first observed on disclosure day or within the next three days, and that 5%-12% of antivirus engines detected them. It adds that some may be incomplete proofs of concept, so this is not evidence of 10 successful compromises.
Am I affected?
Not confirmed. The source does not name affected products, versions or vendors for the PHP2Shell vulnerabilities, so it cannot be used to decide whether a specific system is vulnerable. It also does not describe any confirmed victim.
Points the source does support:
- The LLM API counts are from ReversingLabs file searches. The company says they may partly reflect bundled SDKs and libraries (2026 files averaged about 1.8 endpoint references per file) rather than live model calls.
- ReversingLabs says affected components embedded in third-party software that teams did not build are harder to inspect.
- Fixed versions or patches are not confirmed in the source.
What to do now
- Check your own inventory for exposure to CVE-2026-109618 and CVE-2026-109619, and consult the vendor or NVD records for affected versions, since the ReversingLabs post does not list them.
- Inspect delivered software artifacts, downloaded packages and files before deployment or execution, including embedded components in third-party software. ReversingLabs recommends this.
- For a critical, internet-facing remote-code-execution vulnerability with credible exploitation, ReversingLabs suggests starting emergency mitigation immediately and prioritizing patching or compensating controls within 24 hours where feasible, with asset discovery and exposure validation in parallel.
- Monitor behavior as well as file signatures, including suspicious process activity and unexpected outbound model-API requests from software with no approved reason to use them (ReversingLabs recommendation).
- Give AI agents and integrations only the credentials, network access and execution privileges they need, and prepare to revoke credentials quickly after suspected exposure (ReversingLabs recommendation).