2 November 2026 · AI in public spaces and information

A system can search for a face in a crowd, summarise a news story or use content published online. Today’s briefing connects these three areas through original sources: the results of a police trial, a survey of readers and an announcement about publishers’ preferences. Dates appear alongside each document: this is a morning selection, not a list of announcements all released overnight.

Western Australia · Facial recognition moves to evaluation

In its 2 November update, WA Police reports that the trial ending on 31 October performed 1,380,882 facial scans, generated 292 alerts and resulted in 122 arrests. Nine incorrect alerts are reported. Matching against a watchlist is followed by officers’ verification: an alert alone is not grounds for arrest. An internal evaluation will examine performance, effectiveness and community impact. Scans, alerts and arrests measure different stages and should not be conflated.

WA Police · Facial recognition trial results, updated 2 November 2026 ↗

Reuters Institute · Using AI for news is not the same as entrusting it with journalism

The report dated 29 October finds that, across six countries comparable since 2024, weekly use of generative AI to obtain news rose from 3% to 9% in 2026. The full survey covers nine countries and does not include Italy. Chatbot-mediated access is growing, but acceptance of news produced entirely by AI remains limited. These are survey responses, not automatically measured behaviours or a certification of system reliability.

Reuters Institute · Generative AI and News Report, 29 October 2026 ↗

Web content · A common signal for training preferences

On 30 October, OpenAI announced it would honour a shared robots.txt signal for expressing preferences about content use in training. Implementation is planned over the coming months and existing controls remain valid. This is a voluntary initiative connected to IETF work, not a universal block already active against every AI system. For publishers, the aim is to express a choice to multiple participating providers with fewer separate rules.

OpenAI · Website AI training preferences, 30 October 2026 ↗

The desk’s perspective · Control must be identifiable

The common thread is the ability to identify who controls each stage. In security, an alert requires verification; in information, a summary requires accessible sources; in data use, a preference needs a clear scope and timeline. This is the desk’s interpretation: technological capability and responsibility remain two questions to address together.

Sources and documents

Document date or reference period: Selection of 2 November 2026: WA Police update of 2 November; Reuters Institute report of 29 October and OpenAI announcement of 30 October. These are not all announcements from the Italian overnight period..

Desk selection and summary prepared with AI assistance, based on the three linked original sources. Police-reported results, survey responses and provider announcements are distinguished. The conclusion is the desk’s interpretation.

Report an update or correction →