Compiled by Sarah Willrich.
Articles
Kate Klonick and Renée DiResta highlighted how decisions made by artificial intelligence (AI) companies about what information their models consult impact the information they provide to users. Klonick and DiResta identified three key pathways through which this occurs: a decrease in reliance on some sources, relied-upon sources themselves no longer being updated, and the acquisition of new, exclusive datasets.
What a model learned from, therefore, is not necessarily the same as what it looks up. Paywalls, crawler restrictions, commercial arrangements, and product choices shape AI-generated responses. Both training and retrieval are shaped by consequential forces largely invisible to the public, which increasingly relies on the answers.
We call this phenomenon invisible source skew: the gradual shaping of an information environment by routine product decisions made at AI companies about which sources are retrieved, which are maintained, and even which datasets are acquired. Any one of those decisions may make technical sense, be commercially rational, or even be entirely unremarkable. But together, they create a public epistemic infrastructure whose inputs can change dramatically without users—or often anyone outside the companies—being able to see, understand, or contest those changes.
U.K. Supreme Court Opens Door for Spyware Victims to Sue Foreign States
Siena Anstis, Natalia Krapiva, and Kate Pundyk analyzed the U.K. Supreme Court’s decision that Bahrain does not have sovereign immunity in a case brought by Bahraini dissidents alleging they were targeted by spyware, even though the perpetrators were not in the U.K. The authors explained that the case adds important legal protections for victims of digital transnational repression but raises important questions about how the new precedent interacts with treaty obligations.
While the Supreme Court’s immunity decision is a big step forward for transnational spyware litigation, there will be other challenges. In particular, future claimants will have to establish that their claim falls within the scope of Section 5 of the SIA, which is confined to “death or personal injury” and “damage to or loss of tangible property.” A claimant must plead one of those qualifying heads to fit within Section 5 of the SIA and defeat immunity. As Lord George Leggatt observed in his dissenting opinion in
Shehabi, claims framed solely around economic loss or the exposure of private information—with no pleaded personal injury or property damage—will fall outside the exception. However, once a claimant overcomes this obstacle, consequential loss flowing from that injury is potentially recoverable.
China’s AI-Enabled APT Operations Are Getting Interesting
In the latest edition of the Seriously Risky Business cybersecurity newsletter, Tom Uren discussed a new report about Chinese AI malware capabilities, the Treasury Department’s naming of specific Iranian hackers in its Operation Economic Outcast press release, whether AI should be considered critical infrastructure, and more.
Bitdefender suspects that the same high-level specification document was independently implemented twice with AI assistance. Bitdefender concedes that this structural similarity is not conclusive evidence but notes that it is the “kind of thing an AI-assisted workflow makes easy.”
This is the first example we’ve seen where the evidence tells a compelling story of a competent cyberespionage actor incorporating AI into its work practices. SilkParasite is taking the same, disciplined approach to malware development and doing more of it. It’s creating more malware families to build redundancy, make attribution and discovery harder, and reduce the risk of compromise from any single exposure.
Podcasts
Lawfare Daily: Israel’s Most Consequential Election, with Natan Sachs: Natan Sachs joined Daniel Byman to discuss who is likely to win the Oct. 27 Israeli elections, what the consequences might be for wars in Gaza and Iran, how the winner could impact Israeli domestic politics, the role of small parties, and how Trump might sway the vote.
Scaling Laws: Founders & Founders: Brandon Mitchell of WriteSea: Brandon Mitchell spoke with Kevin Frazier about the growing role of AI in workforce development, how Mitchell’s company provides career guidance in a hiring environment that increasingly relies on automated screening tools, how AI will disrupt the labor market more broadly, and more.
Videos and Webinars
Lawfare Live: The Trials of the Trump Administration, August 28: Benjamin Wittes sat down with Molly Roberts, Anna Bower, and Eric Columbus to discuss the latest developments in litigation surrounding the Trump administration, including updates in the litigation over President Trump’s executive order targeting mail-in voting, the Justice Department’s filing suggesting that the Kennedy Center may be torn down, the prosecution of Rep. LaMonica McIver, and more.
Announcements
*Lawfare *is hiring a Program Assistant for our AI Research Program and seeking applications for the 2026-2027 Student Contributor Program. Learn more and apply at the links above.
*Support *Lawfare
*Follow us on Twitter, Facebook, YouTube, and LinkedIn. Become a material supporter on Patreon. Sign up to receive *Lawfare
