After Anthropic refused to compromise and faced retaliatory sanctions from the U.S. Department of Defense, the Pentagon's AI contracts are rapidly shifting to other willing tech companies. According to...Bloomberg NewsGoogle has officially rolled out its Gemini-based AI Agents service to over 300 million military and civilian personnel at the U.S. Department of Defense. These AI tools will initially be deployed on unclassified networks to handle tedious tasks such as budget preparation and meeting summaries. This collaboration is not only expected to expand to top-secret systems in the future, but it has also reignited Google's sensitivity regarding the weaponization and military applications of AI.
Import 8 exclusive AI agents and enable natural language customization.
According to Emil Michael, Under Secretary of Defense for Research and Engineering, Google's AI agent service will initially run on unclassified networks, but the two sides are already in negotiations to integrate these powerful generative AI tools into Classified and Top-secret systems.
In an official blog post, Google Vice President Jim Kelly pointed out that Google provided the Pentagon with eight pre-built AI agent services specifically designed to automate routine administrative tasks, such as automatically summarizing meeting minutes, compiling budget spreadsheets, and even cross-checking and verifying action plans proposed by the military against the National Defense Strategy.
In addition, Department of Defense personnel can also use natural language to quickly create customized AI agent functions for their own business needs.
The import speed far exceeded expectations, but the training volume is alarmingly low.
In fact, U.S. Department of Defense employees are far more receptive to AI tools than one might imagine.
Since last December, Department of Defense personnel have been able to access Google's AI chatbot for non-classified tasks through a dedicated portal, "GenAI.mil". In just a few months, as many as 12 million employees have used the service, entering more than 40 million unique prompts and uploading more than 4 million documents for AI analysis.
However, the Department of Defense's internal training is clearly lagging behind the pace of technology adoption. Data shows that only 26000 people have completed relevant AI training courses since December; although all future training sessions are fully booked, this "learn by doing" gap still raises concerns about the rigor of internal data processing within the military.
Taking advantage of the situation or a moral compromise? Google's quiet change to its AI principles sparks controversy.
The timing of this large-scale deployment is extremely sensitive. Not long ago, Anthropic, an AI startup known for its security, completely broke with the Pentagon and was even criticized by the military for refusing to remove security safeguards related to "domestic surveillance" and "autonomous weapons" from its models.Label it as "supply chain risk",andTaking it to court.
After Anthropic withdrew, the Pentagon quickly turned to companies like OpenAI, xAI, and Google, signing restricted network cooperation agreements.
This has sparked a huge backlash within Google, with approximately 900 Google employees and 100 OpenAI employees recently signing an open letter strongly urging their employer to uphold its bottom line of not using AI for domestic surveillance and autonomous weapons. Back in 2018, Google terminated its contract with the Pentagon due to strong protests from thousands of employees against "Project Maven," which analyzed drone imagery. Intriguingly, reports indicate that Google "quietly revised" its official AI Principles in early February of this year, significantly relaxing restrictions on its military applications.
