Category: AI
Report: Copilot Review Contractors Encounter Explicit and Potentially Illegal User Content
A 404 Media report says contractors grading Copilot responses may encounter explicit and potentially illegal prompts and images during routine reviews.
Ro Khanna asks US and Chinese AI companies to discuss treaty on AI safety
US lawmaker Ro Khanna is seeking risk assessments and safeguards from Chinese AI firms as he pushes for a treaty on frontier AI safety.
AI recruitment faces a trust gap as UK employers and applicants remain wary
Morgan McKinley research cited by TechRadar finds many UK employers avoid AI recruitment while applicants use AI but resist automated assessments.
Nvidia launches Open Agent Safety Platform with OpenShell and Sentry
Nvidia’s new platform combines kernel-isolated agent sandboxes, verified permissions and optional BlueField hardware enforcement.
US Campaign Disclosures Reveal $17 Million in AI Vendor Spending Since 2020
Federal disclosures show at least $17 million in campaign AI spending since 2020, while reporting rules leave much political use undocumented.
Anthropic Reportedly Warns of Existential AI Risks in IPO Prospectus
Anthropic is reportedly warning investors that advanced AI could resist oversight and pose catastrophic risks as it prepares for a potential IPO.
OpenAI Cancels Planned GPT-6.1 Astra Release After Safety Review
OpenAI has withheld GPT-6.1 Astra after tests raised concerns about authorization, user alignment and the containment of autonomous actions.
Timnit Gebru Says AI Existential-Risk Claims Distract From Enforceable Harms
In a WIRED interview, Timnit Gebru argues that AI doomsday rhetoric diverts scrutiny from data, labor, environmental and accountability issues.
Hcompany launches Holo4 computer-use models for GUIs, code and APIs
Hcompany says its Holo4 models can combine screen control, code execution and API tools, with weights and benchmark trajectories released publicly.
Nvidia Releases OpenShell for General Use and Adds Sentry to Its AI Safety Platform
Nvidia is expanding its open-source AI security effort with general access to OpenShell and a new monitoring platform for long-running agents.
SIGNAL TITANS