Threat researchCisent
AI has a supply chain too: the LiteLLM case
Chile's cyber agency warned some twenty institutions after malicious versions of an open-source AI tool were published.

On August 12, 2026, Chile’s National Cybersecurity Agency (ANCI) activated a preventive protocol and warned some twenty Chilean institutions, including government sites, that they could be exposed by the LiteLLM hack. LiteLLM is an open-source gateway that many companies use to connect their applications to different artificial intelligence models.1
What happened
According to the account published by BioBioChile, citing the firm Hudson Rock, a group called TeamPCP first compromised the automation of Trivy, an open-source vulnerability scanner. LiteLLM’s developers used Trivy in their continuous integration pipeline, and through it the attackers got the keys to publish versions of LiteLLM with malicious code, which quietly stole secrets from the environments where they were installed.1
ANCI clarified that this was a possible exposure, not confirmed incidents, and sent instructions for checking systems and credentials.1
Why it matters
It’s a supply-chain attack with two hops: a security tool compromised an AI tool, and the AI tool compromised the people who installed it. None of the victims did anything unusual: updating a dependency is exactly what everyone recommends.
What does leave a trace is the theft itself. To carry off the secrets, the malicious code has to connect outward, to an attacker’s server, from machines that normally don’t talk to anyone new.
Where Cisent comes in
Sentinel Defense detects, assesses and blocks attacks alongside the organization’s EDR, its firewall and every source it connects. A build machine that suddenly opens a connection to an unknown destination is an intrusive connection, and Defense blocks it on the spot. Every new attacker technique, like this one, becomes a defense rule.
And Cisent works with its own models, with no outside AI provider in between.
Sources
BioBioChile, “ANCI activa protocolo preventivo y notifica a veintena de instituciones chilenas por hackeo a LiteLLM” (in Spanish), August 14, 2026. https://www.biobiochile.cl/noticias/nacional/chile/2026/08/14/anci-activa-protocolo-preventivo-y-notifica-a-veintena-de-instituciones-chilenas-por-hackeo-a-litellm.shtml↩︎