Private cyber power needs public accountability, not just federal approval

The most consequential phrase in Washington’s latest cyber policy may not be offensive operations. It is operational partner.

The FBI’s strategy published on 9 September 2026 places disruption and stronger partnerships among its central priorities. Federal News Network subsequently reported that FBI Cyber Division Assistant Director Brett Leatherman said the bureau wanted industry to move beyond exchanging intelligence with government and become an operational partner. That language indicates a deeper relationship between the state and private technology companies, even though it does not reveal the eventual scale or methods of the programme now being implemented.

The change should not be judged mainly by asking whether companies can conduct cyber operations faster or more effectively than government agencies. The harder question is whether Washington can build publicly assessable accountability around businesses exercising, supporting or enabling coercive state power.

Private expertise may add useful capacity. But capacity is not legitimacy, and a federal approval is not a complete system of accountability. The programme should not be scaled until authorisation standards, oversight, liability, worker protections, foreign deconfliction and remedies for unintended harm are clear enough to withstand scrutiny.

Not ordinary corporate retaliation

President Donald Trump’s 12 August memorandum directed the Homeland Security Task Force’s National Coordination Center to establish a programme for operations against foreign transnational criminal organisations targeting Americans. According to the White House, voluntarily participating and vetted US companies would work under government direction, control and authority. They could gather threat information and propose operations.

This makes the popular hack-back label an imperfect description. The public framework does not describe a victim company independently pursuing an attacker. It requires operations to be conducted on behalf of and under federal supervision, with written approval from programme directors at the Justice Department and Department of Homeland Security. Participating companies must have a contract with one of those departments.

Those constraints matter. They make the proposal more controlled than corporate vigilantism. The framework also bars operations likely to cause death or serious injury, or to cross the international law threshold for a use of force or armed attack.

Yet government supervision does not reduce the work to a routine technology contract. Mayer Brown’s legal analysis says the memorandum contemplates surveillance and cyber effects that could manipulate, disrupt, deny, degrade or destroy targeted systems, infrastructure or data. Those are coercive capabilities. When a company applies or supports them for the state, the relevant standard cannot be the same one used for an ordinary software procurement.

The public text also leaves a foundational question unanswered. Mayer Brown identified ambiguity over whether private employees or government personnel would execute operations. If government officers retain direct execution authority while contractors provide intelligence, tools and specialist support, some concerns about delegating coercive power would lessen. Questions about data governance, employee exposure, liability and customer trust would remain.

The capacity case deserves a fair hearing

There is a serious argument for a controlled experiment. The White House says US consumers reported more than US$20.8 billion in losses from cyber-enabled crime in 2025, although the underlying dataset was not verified in the reviewed material. The FBI’s new strategy is organised around disrupting adversaries, supporting victims, strengthening partnerships and building capabilities. Private firms can possess expertise and visibility useful to that mission.

Government contracts, written approvals and federal supervision should also make this model safer than allowing businesses to retaliate independently. Participation is voluntary, which gives companies the ability to decline work that conflicts with their risk tolerance or customer commitments.

But none of those premises establishes that private offensive operations will be effective or proportionate. As of the research cut-off on 18 September, no participating company or completed operation had been identified in the reviewed public record. The White House fact sheet did not name any participant. Federal News Network reported that agencies were still implementing the programme. Mayer Brown’s review says the programme directors had been instructed to issue operating procedures within 60 days of the memorandum.

The programme is therefore still a governance proposition, not a demonstrated operational success. Advocates can reasonably argue that private capacity is worth testing. They cannot yet point to public evidence that this particular model reduces cybercrime more effectively than existing disruption methods.

Boards would inherit a public power problem

Any company considering participation would face a decision far larger than whether its technical team can complete the assignment. Its board would need to govern state-directed offensive work alongside responsibilities to employees, customers and international operations.

That starts with authority inside the company. Directors should know who can approve participation, who can halt an operation and what happens when the factual or legal basis changes. Employees need defined escalation channels and protection when they raise concerns or invoke stop-work rules. Contracts should allocate liability rather than leaving workers to discover their exposure after something goes wrong.

Voluntary participation does not answer these questions. A company can decline the contract, but individual employees may still face professional pressure once their employer has accepted it. Nor does voluntary entry remove reputational consequences for a provider whose commercial products or services are used around the world.

For Australian and other allied customers, the concern is not evidence that any particular US vendor has joined. None has been publicly identified. It is the possibility that a globally trusted commercial provider could come to be perceived as an extension of US offensive power. That perception could arise even where an operation is authorised under US rules.

The distinction between commercial service and government mission therefore needs more than an internal policy. Participating firms would need credible separation between offensive work, customer systems, commercial products and the data entrusted to them. The reviewed record does not show how such separation would work, so it should be treated as a requirement to resolve, not a safeguard that already exists.

Secrecy cannot mean institutional invisibility

Some operating details must remain secret. Publishing targets, methods or operational timing could help criminal organisations evade disruption. But protecting methods does not require hiding the architecture of accountability.

Washington can disclose the legal basis for the programme, the evidentiary threshold for selecting targets and the bodies responsible for approval and review. It can require logged activity, auditable decisions and after-action investigations without releasing sensitive technical details. It can publish aggregate outcomes, including suspended operations and unintended effects, while protecting operational security.

A defensible model should also specify dual written approval, independent legal review and a clear stop-work authority. It needs procedures for foreign deconfliction, especially where targeted activity touches infrastructure or data outside the United States. Contracts should establish indemnity, insurance and personal protections for assigned workers. Innocent infrastructure owners and customers need a route to notification, investigation and remedy when disclosure is possible.

These are recommendations, not a description of controls already in place. No public operating procedures, contracts or independent evaluation were available for review. Mayer Brown has also identified unresolved constitutional, surveillance, liability and international law questions, but those are issues for analysis rather than settled legal conclusions.

Independent oversight is essential precisely because the government and its contractors will possess most of the relevant evidence. Written approval by the programme’s own directors is a control, but it is not a substitute for scrutiny by a body able to test whether authorisations, operations and remedies match the published framework.

A pilot must test governance as well as capability

If the eventual programme uses narrow targets, documented evidentiary standards, independent legal review, dual approval, comprehensive logging and meaningful remedies, a limited pilot could add private capacity while remaining more controlled than independent retaliation. That is a conditional case, not a forecast. There is no completed operation or public evaluation on which to base a performance judgement.

The opposite scenario is also conditional. If contracts reward operational tempo without equivalent accountability, or if attribution and stop-work standards remain opaque, the model could expose innocent systems to disruption and participating vendors to retaliation or distrust. There is no evidence that those weaknesses currently exist. Their plausibility is why the implementation rules matter before expansion, rather than after harm forces the issue.

Washington is entitled to seek new ways to disrupt foreign cybercrime. It is also right to reject the idea that every company should be free to retaliate on its own. But moving private industry inside a federal chain of command does not settle the legitimacy question. It relocates it.

The proper test is whether the public can assess who authorises this power, who audits it, who can stop it and who provides a remedy when it causes unintended harm. Until those answers exist, operational partnership should remain a tightly bounded experiment. Speed and specialist skill may justify trying the model. They do not justify scaling power that the public cannot meaningfully examine.

Separating technical capacity from public legitimacy leaves a practical learning task: understanding authorised security testing without confusing technical skill with public authority. The Hack Academy’s online cybersecurity training programme offers self-paced learning, pairing theory and practical exercises in cybersecurity fundamentals with ethical, authorised penetration testing in controlled environments.