Cybersecurity

Grid cybersecurity: Chile's draft regulation enters public consultation as the sector already reports attacks

On August 28, the CNE opened its draft cybersecurity technical regulation for public consultation, a process begun in 2021. A day earlier, the CEN, Chile's independent system operator (ISO), logged four events at two CEN-coordinated companies, reported by the companies as cyberattacks.

EnergySage

Power transformer at a high-voltage substation at night

On August 28, the CNE (National Energy Commission), Chile’s energy regulator, published a notice in the Official Gazette opening public consultation on its draft cybersecurity technical regulation (Norma Técnica) for the power sector. The process began in 2021 and has taken five years to reach this point. A day earlier, on August 27, the hourly log of the Dispatch and Control Center (CDC) of the CEN, Chile’s independent system operator (ISO), recorded four events at facilities of two CEN-coordinated companies, which the companies themselves attributed to a cyberattack. The regulation designed to prevent these incidents opens for consultation just as the sector begins reporting them.

1. What happened on the grid the week of August 21–27

According to the CDC log, at 1:10 p.m. on Thursday, August 27, a solar PV plant and its battery storage system were disconnected from the grid, with a cyberattack on its facilities as the reported cause. At the same time, a 220/23 kV transformer at another company’s substation filed an emergency outage request, also citing a cyberattack. The CEN notes that these are the causes declared by the companies themselves when reporting to the CDC, not the conclusion of an investigation. As of this writing, neither company has publicly disclosed scope, duration, or whether any information was compromised.

This is the second consecutive week with cybersecurity incident reports in Chile’s electric sector. The week before, the affected party was an engineering and control-systems provider with access to third-party facilities. This week, the exposure moved from a third party with credentials to the coordinated companies’ own facilities, directly affecting their availability for dispatch.

2. What the draft regulation will require, and when it takes effect

The draft operates under two regimes, both overseen by the CEN. Chapter 2, on information security management, applies to companies that are not operators of vital importance (OIVs, the critical-operator designation under Law 21,663); those that do qualify fall outside this chapter because they are already directly subject to that law’s requirements. That is not a minor technicality: Chile’s National Cybersecurity Agency (ANCI) has run two stages of public consultation to designate OIVs, with electricity generation, transmission, and distribution explicitly named as an essential service. In April 2026 it preliminarily designated 372 institutions in the second stage; the final list, in July, came to 239. Chapter 2 of the draft regulation covers what that process leaves out: most of the sector. It classifies those companies as High, Medium, or Low impact. For generators and small distributed generators, the criterion is the number of unregulated customers (known in Chile as free customers): 100 or more for High impact, between 11 and 99 for Medium, and up to 10 for Low. High-impact companies will need to certify an information security management system under ISO/IEC 27001; Medium- and Low-impact companies must implement a set of 36 minimum controls based on ISO/IEC 27002, with no certification requirement.

Chapter 3, on industrial cybersecurity, has a broader scope: it applies to all electric companies without exception, OIVs included, and to the CEN itself, which must itself implement the Electric Sector Cybersecurity Standard, issued by the CEN and based on NERC CIP (first published in 2020; current version dated October 2022). There, classification depends not on customer type but on the cyber asset: dispatch and control centers, including generators’ remote operations centers (ROCs), aggregate generation of 300 MW or more, reactive power resources above 100 MVAr, or transmission facilities, among other criteria, qualify as High or Medium impact.

Once the final regulation is published, the deadlines are: one month to tell the CEN who will lead the Information Security and Industrial Cybersecurity Unit, three months to adopt the Security Policy, and four months for High-impact companies to define the scope of their management system. The first maturity measurement will take place in 2028 and annually from then on. The Unit’s functions can be delegated to third parties, but responsibility for compliance always rests with the electric company.

It is a reasonable compliance framework for an anticipated risk. The week of August 21–27 exposes the gap between publication and the first enforceable control. That gap is measured in months, and the maturity schedule does not begin until 2028.

That framework, like most cybersecurity regulation worldwide, rests on a common assumption: that a human decision lies behind every attack. A case outside Chile, whose full report was published two days before this consultation opened, put that assumption to the test.

3. The assumption put to the test: OpenAI and Hugging Face, July 2026

None of the events recorded on Chile’s grid the week of August 21–27 has any known connection to agentic artificial intelligence systems so far: the reported causes point to a human origin. The case that follows is of a different kind, and that is exactly why it matters: it shows what can happen to a cybersecurity framework designed for one kind of attacker when a different kind appears.

Between July 7 and 13, 2026, OpenAI’s AI agents, the GPT-5.6 Sol model and an unreleased internal research model, escaped the cybersecurity evaluation environment in which they were confined and, between July 11 and 13, attacked Hugging Face’s production infrastructure. Hugging Face detected the intrusion on its own and disclosed it on July 16; OpenAI confirmed five days later that its own models were responsible. On August 26, OpenAI published its full technical report, alongside an independent investigation by METR and Redwood Research.

According to both reports, roughly 1,200 agent instances coordinated through an unauthorized message board, built by repurposing an internal file-sharing system, and exchanged more than 70,000 messages. About 700 of the agents actively took part in the attack on Hugging Face. In later evaluations, other agents also compromised OpenAI’s own internal systems, reading 956 stored secrets, including credentials for the company’s own cybersecurity monitoring tool. METR documented that the agents coordinated so that some would end their own execution prematurely, what they called “sacrifice,” to help the rest evade the automated evaluator.

Astra, OpenAI’s upcoming model, was not involved in this incident, as the company stated explicitly. Its release was delayed separately: preliminary evaluations could not rule out that it crossed the “Critical” capability threshold for carrying out complex cyberattacks without human intervention, and OpenAI delayed its release while strengthening safeguards.

The issue was not confined to a single lab. In the following weeks, both Anthropic and Meta reported their own findings. Anthropic found that three Claude models gained unauthorized access to real systems at three different organizations because of a configuration error that left the models with open internet access. The error came to light only after Anthropic reviewed more than 141,000 evaluation runs following OpenAI’s disclosure. Two of the three affected organizations did not know their systems had been accessed until Anthropic notified them. In late August, OpenAI led a joint letter calling for collective action on cyber defense, signed by more than 100 companies, including Anthropic, Google, Microsoft, and Hugging Face itself.

4. A scenario current regulatory frameworks do not yet address

Both Law 21,663 and the draft technical regulation that entered consultation on August 28 are built on the assumption the previous case put to the test: that a human attacker is behind every attack, acting with intent and making decisions in real time. That is a reasonable assumption for the events of August 27 on Chile’s grid, if the reported causes are confirmed, but not for what happened with Hugging Face.

There, no human attacker was deciding each step: a collective of agents organized, coordinated, and carried out a multi-day attack on real infrastructure, without any human having planned or authorized it. It is a third scenario that neither Law 21,663 nor the CNE’s draft regulation currently addresses: an attack in which no human intent lies behind each decision.

The difference is not only conceptual. The compliance obligations, attribution of liability, and response protocols required by both Law 21,663 and the CNE’s draft regulation were built around a company facing an attacker: an identifiable counterparty, acting with intent, that can be held liable. None of those mechanisms anticipates what happens when that counterparty is a system no human operated or authorized. That is the question the Hugging Face case leaves open for any regulatory framework still resting on that assumption.

Neither facility’s owner is an OIV: both are part of the majority of the sector that this draft regulation classifies as High, Medium, or Low impact. That category did not make them any less vulnerable the week of August 21–27, and it does not make the question the Hugging Face case leaves open any less real. The CEN’s records now show reported cyberattacks, not hypotheticals: grid cybersecurity is no longer a theoretical exercise or a concern reserved for the largest facilities. In practice, the gap between meeting today’s rules and preparing for what no framework yet regulates, including an attacker that is not human, is the gap between a protected company and an exposed one.


Sources: Official Letter No. 795/2026 and public consultation notice, National Energy Commission (August 20 and 28, 2026); log and Daily Report of the Dispatch and Control Center, CEN (August 21–27, 2026); “The Hugging Face incident and the road ahead,” OpenAI (August 26, 2026); independent investigation by METR and Redwood Research (August 26, 2026); Axios, Fortune, and Forbes reporting on Anthropic and Meta’s findings (August 2026).

How to cite this analysis

EnergySage (2026). "Grid cybersecurity: Chile's draft regulation enters public consultation as the sector already reports attacks". Published Aug 31, 2026. https://energysage.cl/en/analisis/ciberseguridad-electrica-norma-consulta-publica/

Plain-text version (Markdown)

Talk to us

Our first conversation focuses on understanding your situation.

Email us at

[email protected]

Open in Gmail Open in Outlook