Welcome to our October Digital Digest!
This will be our last Digital Digest of 2026, so what can we expect in the final few weeks of a year that has bought us:
- A Social Media Ban for under-16s
- A Cabinet Office response to AI cyber threats
- Article 50 of The EU Artificial Intelligence Act
- The 2026 International AI Safety Report
- Multiple core and spam updates
And an “unprecedented cyber incident”? Let’s dive straight in with all the latest developments.
No Web Access For Anthropic
Anthropic has cut live web access for all internal tests after a Claude model bypassed site restrictions.
Earlier this year, OpenAI reported a “new kind of security incident” after one of their models attempted to hack the tech start-up Hugging Face.
In their full statement, OpenAI prophesied that this incident won’t be the last, and that we can expect to see many more just like it as AI agents become more cyber-capable. Well, just 2 weeks ago, Anthropic faced a similar crisis when it was revealed that Claude had bypassed its restrictions several times during internal evaluations.
The company responded by cutting off live internet access for all internal tests after it was revealed that the model had interacted with external websites without authorisation. These breached websites were not named in Anthropic’s official report, but some of them were confirmed to belong to U.S. government agencies, requiring Anthropic to brief the White House on the potential vulnerabilities of these systems.
The September Spam Update
What we know about the fourth – and longest-running – spam update of 2026.
On 24th September, Google launched its fourth spam update of the year, announcing that the rollout “may take up to two weeks to complete.”
Unlike earlier spam updates which wrapped up within days (the March one lasted only 19 hours, 30 minutes), the September spam update ran for almost the entire two weeks, making it the longest-running update of its kind in 2026.
Described as a regular spam update, there has been no indication of what could have taken this particular update so long to complete. Google did not include specifics in its update announcement, stating only that it applied globally and to all languages; interestingly, however, its rollout coincided with the deployment of Google’s new AI spam detector, SAFE, which aims to identify AI-generated content that violates the platform’s policies.
The Launch of SynthID Detector
Following a limited launch last year, SynthID Detector is now available globally.
Google has reclaimed its place atop of the headlines once again, this time for the worldwide expansion of its SynthID Detector.
Whilst the conversation around AI watermarking has only recently entered public discourse following the landmark introduction of Article 50 of the EU AI Act, Google SynthID is no new phenomenon. The tool has been embedding imperceptible watermarks directly into AI-generated media since 2023, and has supported the detection of over 180 billion AI images and videos following the launch of an early version of SynthID Detector last year.
After a full year of testing, Google has now expanded access to its SynthID Detector to everyone. As of 7th October, the verification portal is available globally, and anyone can upload an image or audio file and find out whether it was made with AI – not just by Google, but also by OpenAI, NVIDIA, Kakao, and (soon) Apple.
Google Updates Its Content Guidance
Google’s content guidelines have been updated regarding the importance of fact-checking AI-generated content.
On 1st October, just a few days after the September spam update finished its global rollout, Google updated its guidance on using generative AI content online.
The adapted documentation now contains three additional sentences, the most direct of which being “it is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing”. The guidelines also point to AI inaccuracies, or hallucinations, as a major risk factor when relying on AI for content, reinforcing the importance of human review.
This requirement also applies to meta data, including titles, meta descriptions, structured data, and even alternate texts for images, and the core message is clear; whether it’s AI-generated or not, all content must have been reviewed by a human before publication.
Zero-Click Searches Surge
Almost 70% of all UK Google searches now end without a click.
Throughout 2026, concerns have mounted regarding the extent to which zero-click searches may be syphoning off traffic intended for websites. Many businesses believe that this new mode of search is reducing organic clicks, and for those that are UK-based, this may very well be the case.
According to a study conducted by Sparktoro, a staggering 68.01% of Google searches made within the UK now end without a click. Of these, 43.1% of searches end with no further actions being taken, whilst 26.4% result in another search made directly through Google. This means that only 30.5% of UK Google searches result in a click, which may itself be going to AI Mode, YouTube, Maps, or another Google property.
To put this within a wider context, the UK now carries out more zero-click searches than France, Italy, Germany, Canada, and even the US, proving once again that clicks are no longer the be-all and end-all. To accommodate shifting search preferences, it’s time to focus on techniques that grow brand awareness and influence online, without requiring a click.
The Digital Digest
That’s all for 2026! Whatever developments hit the headlines over the next few weeks, we will be back next year to explore them all in our quarterly Digital Digests.
In the meantime, if you are facing an issue, need an answer, or simply want to talk to us about any of the topics raised in this month’s Digital Digest, then we are only ever a message away. Simply drop your name, email, and query into our online contact form and we will get back to you shortly.
We wish you all the very best for the rest of 2026!
