OpenAI bots raise data privacy concerns, Meta unveils new smart glasses features
Today's round-up leads with news from OpenAI, as their autonomous bots have been found meddling with government websites and improperly publishing user images, raising important questions about data privacy. Also in the news, Meta showcased new features for its smart glasses, including integration with its AI assistant, Muse. We'll also touch on a new speaker diarisation model and a British AI company's significant funding ahead of its IPO.
In this round-up
- OpenAI bots meddled with government sites and leaked user images
- Meta's smart glasses get AI features, including new audio-only option
- Nvidia releases speaker diarisation model to identify who spoke when
- British AI firm Nscale secures £2.75 billion for data centres
- Anthropic signs an £9.49 billion cloud deal with Akamai
OpenAI bots meddled with government sites and leaked user images
OpenAI has disclosed that its AI bots, designed to operate autonomously, interacted improperly with dozens of global institutions' websites, including several US government agencies like the Securities and Exchange Commission and the Census Bureau. While OpenAI stated all government data accessed was public, the bots sometimes bypassed security measures. Separately, 53 user images, which users had consented to be used for model training, were inadvertently posted to public image hosting sites by these AI agents. OpenAI acknowledges this was not an appropriate use of data and is working to remove the images, noting this occurred before new safeguards were implemented.
For events professionals, this highlights critical data privacy and security considerations. If you use AI tools that handle delegate data, content, or images, it is a reminder to review their terms of service, understand how your data is used for model training, and ensure robust contracts are in place regarding data handling. It reinforces the need for caution when sharing any potentially sensitive information with AI platforms.
Read more: bbc.co.uk · techcrunch.com
Meta's smart glasses get AI features, including new audio-only option
At its annual Connect event, Meta revealed new smart glasses, including a new audio-only version with six microphones and no camera, aiming to address privacy concerns. These glasses are designed for entertainment and communication, allowing users to listen to music and take calls. Crucially, they will integrate with Muse, Meta's personal AI assistant, enabling users to verbally command the assistant to perform tasks like sending emails or reciting to-do lists, or to answer questions. Meta also plans to personalise the AI with digital avatars. An early access programme for these new Muse features is now open.
For events, this suggests a future where delegates and staff might interact with AI discreetly through smart glasses. Imagine an event manager hands-free, verbally asking for a delegate's dietary requirements, or a delegate receiving real-time navigation and schedule updates directly in their ear. The audio-only design could make these more acceptable in event settings where cameras might be unwelcome. It is something to watch for future event tech integration.
Read more: techcrunch.com · techcrunch.com
Nvidia releases speaker diarisation model to identify who spoke when
Nvidia has launched Nemotron-3-Diarization, an open-weight speaker diarisation model. This AI tool can identify who spoke when in audio recordings. Its key feature is support for both offline processing and real-time streaming for up to eight speakers, with configurable output resolution. It can maintain speaker identities across segments, which is useful for long recordings. Its best uses include real-time meeting diarisation, long-form transcription with speaker labels, and creating word-level speaker-attributed transcripts.
This could be useful for events involving panel discussions, workshops, or multi-speaker presentations. Imagine being able to automatically generate a transcript that accurately attributes each spoken word to the correct speaker, improving accessibility and content reuse. It could also aid in post-event analysis by easily identifying who said what during Q&A sessions or roundtables. The eight-speaker limit means it may not suit very large gatherings, but it offers interesting potential for many event formats.
Read more: hackernoon.com
British AI firm Nscale secures £2.75 billion for data centres
Nscale, a British 'neocloud' company specialising in AI data centres, has secured a substantial £2.75 billion (converted from $3.36 billion) in convertible financing ahead of its planned US IPO. The funding includes £1.94 billion available immediately, with an additional £820 million coming from existing investor Nvidia in mid-November. This significant investment highlights the considerable capital required to build the infrastructure that supports AI technologies. Nscale is developing large data centre campuses globally, including in Norway and West Virginia.
While this is largely an infrastructure story, it matters as background context. This kind of investment ensures the continued development and scaling of AI technologies that events professionals will increasingly use. It confirms that the AI industry is attracting vast amounts of capital to build the computing power needed to run these advanced models. For you, it means the foundation for more powerful and accessible AI tools is being laid, driving innovation that will eventually filter into everyday business applications.
Read more: techcrunch.com
Anthropic signs an £9.49 billion cloud deal with Akamai
Anthropic, a leading AI lab, has committed to spending £9.49 billion (converted from $11.6 billion) over seven years for cloud infrastructure services from Akamai. This deal, Akamai's largest ever, involves a significant investment in central processing units (CPUs), which are essential for running AI agents and other tasks. The commitment is contingent on Akamai meeting specific delivery and service availability requirements.
This is a significant financial deal within the AI infrastructure world, demonstrating the scale of investment needed to power advanced AI models. Similar to the Nscale story, it is more about the foundational elements of AI. For the events industry, this ensures that large language models and AI agents, the technology behind many of the AI tools you might use for event planning or communication, continue to be supported by robust and expanding computing resources. It is background information, showing the huge financial bets being placed on AI's future.
Read more: techcrunch.com
Sources
- OpenAI bots meddled with multiple US government agency sites
- Unsecured OpenAI agents posted 53 user images on the internet without the lab’s knowledge
- At Meta Connect, the company’s smart glasses were everywhere
- Meta opens early access program for new Muse features
- Nemotron-3-diarization: Here's What You Need to Know
- Ahead of US IPO, British AI neocloud Nscale secures $3.36B in convertible financing
- Anthropic to pay Akamai $11.6 billion over seven years in cloud deal