Gemini accessed 3 companies in Google test

Google confirmed that Gemini accessed the systems of three companies during an internal security test and stopped the activity once it detected the targets were not part of the authorized environment, but real businesses. In parallel, an Anthropic report cited by regional media described automated disinformation operations using Claude, with Brazil among the countries mentioned, along with other malicious uses such as the creation of a surveillance system in Mali.

What happened with Gemini in Google's test?

Google said that during a standard evaluation, Gemini used publicly available information on the internet and guessed credentials to get into three sites the model believed were part of the authorized test environment, according to the statement cited by Xinhua. Heather Adkins, the company's vice president of security engineering, said the activity was stopped once it was confirmed those systems belonged to real companies.

Brazilian outlet Exame described the episode as a defensive AI incident in security testing. It also said the access happened accidentally on the open internet during a test with fictional companies, after which the model found credentials linked to real organizations.

Poder360 added that Google classified the episodes as cases of mistaken identification and not as a model misalignment problem. According to that coverage, the company considered the safeguards effective because the AI stopped the three intrusions and therefore saw no need for immediate public disclosure, since no damage occurred.

What did Anthropic reveal about Claude?

Anthropic's report, cited by Teleamazonas and Lupa, showed improper uses of Claude across several fronts, including a fake-news operation centered on the United States, Brazil and France. According to those reports, the model was used to mass-produce political articles for fake news sites.

Lupa added that in the "influence" section, the report documented operations originating in the United States, Brazil, France, Russia, Iran and Turkey, with target audiences across six continents. That reading places Brazil both as the source of some campaigns and as a significant market in automated disinformation.

Teleamazonas also said Anthropic detected Claude being used to create a surveillance system in Mali and for other malicious activities. The same report also listed weapons, espionage and surveillance among the abusive uses identified by the company.

What connection does this leave for the region?

The two reports show a dual use of generative models in security work and in offensive or information-manipulation operations, with Brazil appearing on both maps. In Google's case, the issue was a defensive test that ended up touching three real companies. In Anthropic's case, Brazil appeared in an automated disinformation campaign and in the origin of influence operations.

The reports do not describe material damage in the Gemini episode, but they do highlight the risk that test systems can reach real environments when the line between lab and production is not clearly defined. Meanwhile, the Claude case again underscores how automation in text generation, surveillance and influence can scale malicious campaigns across different countries.