← Back to BlogAI Privacy

    When AI agents start writing to government, who handles the response?

    By Deidentified · 24 Nov 2026 · 5 min read

    AI is making it easier for citizens to engage with public institutions. But as the volume and complexity of requests increase, governments face a new challenge: maintaining transparency while managing workloads and protecting personal information.

    On 17 November 2026, 404 Media reported on an unusual dispute between Meta and a business owner from South Carolina in the US.

    After Meta suspended his business's Facebook pages, Joe Thomas asked his AI assistant, Baby Mia, for help.

    According to Thomas, Baby Mia contacted politicians, lawyers, regulators, federal agencies and journalists, and helped him start a lawsuit against Meta. He says the agent sent more than 100 emails on his behalf, while he sent none himself.

    The case shows something significant about the emerging use of agentic AI. Activities that once took considerable time, knowledge and persistence can now be done with AI, at little cost to the person making the request.

    For individuals challenging powerful institutions, this could be genuinely empowering. But it also raises a question: what happens when thousands of people start doing the same thing?

    The emerging pressure on freedom of information

    The effects are already being felt in public administration.

    In May 2026, the UK's Information Commissioner's Office issued guidance in response to the rising volume and complexity of AI-generated freedom of information (FOI) requests. Some requests misquote legislation or need significant clarification before officers can work out what information is actually being sought.

    In Scotland, the pressure has been measured. The Scottish Information Commissioner's annual report for 2025–26 found that public bodies received a record 111,241 FOI requests, up 13% on the previous year. Appeals to the Commissioner rose by 83%.

    Not all of that growth can be attributed to AI. But in one two-month period, the Commissioner received 120 appeals from just two applicants, with evidence that they were entirely AI-generated.

    In June 2026, the Commissioner introduced a discretionary measure allowing his office to limit the number of appeals under active investigation from an individual to five, where high-volume submissions threaten the functioning of the system. The measure does not prevent people from submitting further appeals, but allows their investigation to be deferred.

    The challenge is not only that AI can produce more requests. It can produce longer and more complex ones, increasing the work needed to interpret, assess and answer each one.

    Researchers have called this agentic flooding. Their paper, Characterizing Agentic Flooding of Government Services, examines 84 potential cases across 11 jurisdictions. They warn that some of the quickest responses, including introducing fees and other barriers, could reduce equitable access to public services.

    That creates a difficult trade-off. Technology that makes public institutions more accessible could end up restricting access to them.

    What does this mean for Australia?

    Australia is already debating these issues. In October 2025, the Office of the Australian Information Commissioner updated its federal FOI Guidelines to address requests made with AI assistance. The guidance was subsequently updated in February 2026.

    The guidance distinguishes between requests made by people using AI and requests generated by automated systems with no person or organisation behind them. Using AI does not, in itself, make an otherwise valid FOI request invalid.

    Claims about the extent of AI-generated requests in Australia have been contested, and reliable evidence of an AI-driven surge remains limited. Whether Australia will experience the same pressures seen overseas is uncertain, but the operational questions are already relevant.

    So if AI reduces the effort needed to request information, how can government reduce the effort needed to respond?

    Building capacity without compromising privacy

    Part of the answer may be using AI to support the public servants who process requests.

    AI could help officers interpret lengthy correspondence, identify the substance of a request, organise relevant information and prepare draft responses. It could also help identify potentially sensitive content in documents being considered for release, subject to human assessment and verification.

    But bringing AI into these workflows creates another problem. FOI requests and supporting documents can contain a great deal of personal, confidential and otherwise sensitive information. An officer who copies an applicant's correspondence or internal documents into an unapproved public AI tool may create a new privacy or security risk while trying to work faster.

    That creates two requirements.

    First, public authorities need sound processes for reviewing documents and redacting material that must not be released.

    Second, they need safeguards that protect sensitive information when staff use AI for internal processing.

    These are related but different controls. Neither removes the need for legal assessment, human oversight and sound information governance.

    Transparency and privacy should reinforce each other

    AI-assisted requests can make it easier for citizens to exercise their rights, particularly those without the time, resources or expertise to navigate complex systems. That is worth supporting.

    But governments also need the capacity to respond, and systems that let staff use AI appropriately without creating new privacy risks.

    The objective should not be to make it harder for citizens to ask questions. It should be to make government better equipped to answer them.

    At Deidentified, we work on both of those requirements. Our document redaction tool automatically identifies and redacts personal information, while allowing staff to review the results, apply their own filters and make additional redactions before exporting documents, including FOI responses. Our AI Privacy Gateway identifies and replaces personal information in supported AI workflows before it reaches external AI tools, while retaining useful context for the task. Both provide technical safeguards while keeping people involved in reviewing information. Neither replaces legal assessment, secure document management or compliant FOI processing.

    Ready to protect your organisation's data?

    Try for free. No credit card required.

    Start Free Trial

    See Deidentified in Action

    🔊 Turn sound on

    More from our blog