Loading story…
MITRE rejected a batch of CVE reports affecting SQLite, libraw, and ESP32-audioI2S after researchers found the underlying claims were likely fabricated. An oss-sec post cited JFrog's review of recently published SQLite advisories and said some records had briefly appeared on cve.org before being withdrawn, underscoring how CVE Numbering Authorities often depend on submitter accuracy and may not be able to independently validate every report before publication.
JFrog said six SQLite entries—CVE-2026-51302, CVE-2026-51303, CVE-2026-51300, CVE-2026-51297, CVE-2026-51296, and CVE-2026-51304—contained major inconsistencies, including references to nonexistent functions, invalid SQL, impossible code paths, mismatched line numbers, and proof-of-concept payloads that did not reproduce crashes. The researchers traced the advisories to a newly created GitHub repository and concluded they were largely AI-generated "LLM slop"; in a broader audit of 55 advisories from the same account, they found 54 appeared fabricated and one described a real bug but with unverified CVE metadata, raising concerns that false high-severity entries can pollute NVD and CISA enrichment pipelines and mislead automated vulnerability triage.
Mallory correlates global threat intelligence with your attack surface — know if you’re exposed before adversaries strike.
5 events from the most recent confirmed update back to the earliest known activity.
Alan Coopersmith posted to the oss-sec mailing list on July 31, 2026, summarizing the rejected CVE reports and citing JFrog's findings on the SQLite entries. The post also noted that CVE Numbering Authorities often rely on submitter honesty because they may lack resources to independently verify every report.
On July 31, 2026, JFrog published an analysis of the SQLite CVE batch after testing the claims against official source code and proof-of-concept payloads. The researchers concluded the advisories were largely fabricated "LLM slop," citing nonexistent code references, invalid patches, and PoCs that failed to trigger the alleged bugs.
JFrog said it reported the fabricated CVE findings to GitHub Security Advisory, Red Hat, and NVD, after which Red Hat and NVD flagged or removed the affected CVE entries. GitHub had not removed the repository at the time of reporting.
MITRE rejected the entire set of questioned CVE entries referenced in the reporting, with the rejection reflected in a CVEProject cvelistV5 GitHub commit. This affected the disputed advisories tied to SQLite, libraw, and ESP32-audioI2S.
A newly created GitHub repository, programmervuln/cveadvisory, published a batch of vulnerability advisories, including many targeting SQLite as well as libraw and ESP32-audioI2S. Some of these CVEs were subsequently published on cve.org earlier in the same week.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
7 references tracked. Mallory keeps watching after this page renders.
malware.news
Open sourcetruesec.com
Open sourcescworld.com
Open sourcetheregister.com
Open sourcereddit.com
Open sourceresearch.jfrog.com
Open sourceseclists.org
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
GNOME's security team changed its vulnerability-handling process by cutting the default disclosure deadline for new reports from 90 days to 30 days and by declining to forward reports to projects that prohibit AI-generated submissions. The policy shift, discussed widely on the `oss-sec` mailing list, reflects the view that most embargoed fixes are completed within the first month and that AI-assisted reporting has become common enough that many submissions now contain at least some machine-generated material. Participants also noted that some projects, including the Linux kernel, have moved away from embargoes for AI-generated reports because such findings can be easily reproduced and redistributed. The change triggered a broader dispute over open-source security governance, with maintainers arguing that low-quality, verbose, or impractical AI-assisted reports are consuming scarce triage capacity and pushing burnout onto already overstretched teams. Mailing-list contributors said many externally submitted bugs are real defects but not necessarily exploitable vulnerabilities, while others warned that rejecting accurate reports could still increase user risk. The debate expanded to the Linux Foundation's **Akrites** initiative, described by critics as an attempt to coordinate AI-era vulnerability response and potentially serve as a "maintainer of last resort" for abandoned critical packages, though others stressed that such a role would not override project ownership or maintainer authority.
Aug 6, 2026
Research from 1Password’s Off-by-1 Labs found that frontier LLMs frequently produced unsafe or incomplete fixes when asked to patch newly disclosed software vulnerabilities. In a study of **6,080 AI-generated patches** for six recent open-source CVEs, only **26.0%** fully resolved the issue without materially changing application behavior; **20.1%** fixed the flaw but altered behavior, while **53.9%** either failed to remediate the bug, introduced a new vulnerability, or both. The researchers said many outputs looked plausible and even passed tests, yet still left exploitable paths open or created fresh defects; in one `freenginx` case study, 114 patches judged to close the original flaw all introduced a new problem. 1Password said expert validation remains necessary and released tooling, datasets, and a paper to support large-scale patch evaluation. The findings coincided with a tougher stance from Linux wireless maintainer **Johannes Berg**, who said AI/LLM-generated patches for that subsystem will be ignored unless a brief review shows they are obviously correct. Berg argued that such fixes often make narrow local changes without adequately accounting for code semantics, architecture, error ordering, or long-term maintainability. Although it was later clarified that `syzbot` AI-assisted patches are pre-reviewed by humans rather than automatically submitted, AI-assisted patch submissions from `syzbot` are being disabled for the Linux wireless codebase, underscoring growing concern that machine-generated fixes can add risk instead of removing it.
Aug 7, 2026
Several **high-severity vulnerability disclosures** were published across widely used developer and infrastructure components, with impacts ranging from **remote code execution (RCE)** to **account takeover** and **arbitrary host file writes**. In *Gogs* (self-hosted Git service), three CVEs were reported: **CVE-2025-64111** (CVSS 9.3) enables RCE by bypassing checks in `UpdateRepoFile` to modify `.git/config` via the API (described as an insufficient fix for an earlier issue); **CVE-2025-64175** (CVSS 7.7) allows a **cross-account 2FA recovery-code bypass** in versions `0.13.3` and earlier if an attacker already has a victim’s username/password; and **CVE-2026-24135** (CVSS 7.2) is a wiki rename path traversal that can delete arbitrary files by manipulating `old_title`. Separately, *Jinjava* (HubSpot CMS template engine) disclosed **CVE-2026-25526** (CVSS 9.8), a sandbox escape chain that permits arbitrary Java code execution by abusing `ForTag` iteration behavior (Bean ELResolver restriction bypass) and `ObjectMapper`-based JSON deserialization to instantiate disallowed classes. A critical Kubernetes storage issue was also disclosed in *Kubernetes Local Path Provisioner*: **CVE-2025-62878** (CVSS 10.0) allows directory traversal via the `parameters.pathPattern` setting, enabling a user who can create storage resources to provision volumes in arbitrary host locations (e.g., `/etc`) and potentially overwrite sensitive files on cluster nodes. In parallel to these product flaws, separate research reported widespread **exposure of Git metadata** on the public internet—approximately **4.96 million** IPs with accessible `.git` directories and **250,000+** exposing `.git/config` files that may contain deployment credentials—highlighting a common, high-impact misconfiguration pattern that can enable source code reconstruction and secret theft. Active exploitation activity was reported for *Ivanti Endpoint Manager Mobile (EPMM)* involving **CVE-2026-1281** and **CVE-2026-1340**, where attackers were observed dropping `/mifs/403.jsp` and using a Base64-delivered Java class loader designed for delayed, in-memory activation rather than immediate interactive webshell use.
Jun 29, 2026