UK intelligence agencies lobbying the government to weaken surveillance laws for AI, report

GCHQ, MI5 and MI6 want safeguards governing the use of bulk personal datasets to be reduced

UK intelligence agencies, including GCHQ, MI5 and MI6, are seeking changes to surveillance laws that they feel restrict their ability to train AI models using extensive sets of personal data.

As reported by The Guardian, spy agencies are currently engaged in lobbying efforts to persuade the government to ease restrictions on their access to specific types of data. They are seeking to relax safeguards intended to protect privacy and prevent the misuse of sensitive information.

Specifically, they want the safeguards governing the use of bulk personal datasets (BPDs) to be reduced. These datasets typically encompass information concerning very large groups of individuals, most of whom are unlikely to be of direct interest to the intelligence services.

MI5, MI6 and GCHQ regularly use BPDs that are sourced from various closed and open channels, and they may also acquire these datasets through covert methods.

The existing safeguards concerning the use of BPDs require the approval of a judge before examination and retention of such datasets can take place.

The intelligence agencies assert that BPDs encompass publicly available data, such as content from video-sharing platforms, public records, podcasts, academic papers and company information, and people have "low or no expectation of privacy" regarding such publicly accessible data.

They say these datasets play a crucial role in identifying potential terrorists and informants, aiding their efforts in national security and counterterrorism operations.

According to The Guardian, the changes proposed by the agencies were presented to David Anderson, a senior barrister and member of the House of Lords.

The Home Office commissioned Anderson earlier this year to conduct an independent review of the modifications to the Investigatory Powers Act.

In his findings, Anderson stated that the agencies' proposals would replace the current safeguards with a faster process of self-authorisation.

Anderson mentioned that the intelligence agencies had been using AI for many years and had already been training machine-learning models using BPDs.

He said substantial growth in the type and quantity of datasets has led to machine learning tools becoming increasingly valuable and beneficial to British intelligence.

According to Anderson, the data obtained from video-sharing platforms, academic papers, podcasts, public records, and company information could be grouped into a new category of BPDs.

The cross-bench peer recommended amending the law to establish "a less burdensome set of safeguards" specifically tailored for this new category of BPDs.

He asserted that the proposed changes would have a relatively minor deregulatory impact.

However, he suggested retaining some level of ministerial and judicial oversight in the process, saying he was opposed to solely relying on intelligence officers to determine which BPDs should be included in the new category.

Privacy experts and civil liberties groups have expressed concerns about the proposed reduction in safeguards.

They argue that as AI capabilities advance, there is a need for stronger regulations rather than weaker ones, and that it should not be made easier for the state to store data on individuals who are not suspects.

In response to Anderson's review, privacy and surveillance expert Ian Brown, expressed his thoughts on his website.

The disappointment of GCHQ data scientists about not being able to freely explore all the exciting possibilities of AI is not a valid justification for weakening fundamental rights protection, he said.

Brown's concerns also extended to the amount of time officials are dedicating to "directly delivering BPD authorisations" to facilitate the intelligence agencies' access to new datasets.

The UK government is currently reviewing Lord Anderson's recommendations and is expected to respond later this year.

The government's response will shed light on how it plans to strike a balance between intelligence needs and safeguarding individual rights and privacy.