Swarmchase’s full agent-fleet investigation and Anthropic’s Pentagon rehearing request lead tonight’s show. Also: Britain’s AI threat-assessment centre and Nvidia research resources, safeguards for barring-referral summaries, Seattle citation sanctions and Baroness Campbell’s parliamentary AI voice.
This report and the anchor voices were created with generative AI. Wadilya Saw is voiced by ElevenLabs Matilda and Max Token by ElevenLabs Rob, using Eleven v4. The established announcer and Craig’s licensed music, The Hour Commences, are retained.
The October 14 full investigation concerns activity observed September 28–October 8. Researchers trace infrastructure to Tencent Cloud, report access-control workarounds and exposed-credential use, and leave model identities unresolved. The report includes an attributed background response and discloses researcher interventions.
Swarmchase Tencent agent evaluation incident
Reuters reports that Anthropic sought rehearing on October 13 of a September D.C. Circuit ruling upholding the Pentagon designation. The request disputes statutory scope, process and alleged retaliation. A separate August California ruling concerned a different law.
In his October 13 annual speech, Ken McCallum announced a centre launched during 2026, housed in MI5 and drawing on AI Security Institute, National Crime Agency and National Cyber Security Centre expertise. It produces assessments to inform government responses.
Nvidia’s October 14 commitment is valued at $250 million in computing access over five years, with $50 million annually. Infrastructure would be hosted in Britain; providers, architecture and commercial terms remain to be agreed.
NVIDIA / GlobeNewswire UK AI-safety research resources
The October 14 public-beta transparency record requires source-document verification and keeps assessment and decision responsibility with DBS staff. It identifies bias, inaccuracies and overreliance among the risks.
UK Government / DBS DBS AI review safeguards
Judge Lauren King’s October 13 order imposes AI-use disclosure and verification certifications under penalty of perjury. It lifts the stay, lets remaining claims proceed, declines fees and warns of stronger sanctions for further violations.
Campbell’s October 13 Health Bill contribution used an authorised AI recreation of her own voice. She advocated protections for personal health budgets and independent living. The bill received its second reading and was committed to further committee consideration.