Cloudflare reveals AI bots dominate web traffic, Google glitches with Gemini Troubleshooting mode, OpenAI upgrades ChatGPT memory, AI CEOs demand Congress tighten bioweapons safeguards

**Carly Wilson** (0:05)
Welcome to Digimasters Shorts.

**Adam Nagus** (0:07)
We are your hosts, Adam Nagus and Carly Wilson, delivering the latest scoop from the digital realm.
The CEOs of Anthropic, OpenAI and Microsoft AI have joined forces to urge Congress to tighten safeguards against fire weapons enabled by artificial intelligence. In an open letter, Dario Amadei, Sam Altman and Mustafa Suleiman called for mandatory screening of synthetic DNA and RNA sales, warning that rapidly advancing AI could erode barriers that once kept biological weapons out of reach. The letter, organized by the Foundation for American Innovation and the Institute for Progress, was also signed by Life Sciences and national security experts, along with several companies that manufacture synthetic genetic material.
The group wants Congress to require industry-wide screening of customers and orders, and to mandate record-keeping to aid potential biosecurity investigations. While voluntary screening already exists, the signatories argue that legal requirements are necessary as generative AI tools spread Pseura and reach a global audience at unprecedented speed. Recent research shows AI adoption has outpaced both the personal computer and the Internet, and publicly available models have demonstrated the ability to provide information about biological weapons. Although biological attacks are historically rare, experts warn they pose unique dangers because agents can be odourless, invisible and highly contagious. Anthrax, for example, can be nearly 100% fatal when inhaled without treatment, as seen in the deadly 2001 attacks that rattled the nation. Existing laws criminalize the development and possession of biological weapons, but lawmakers are now considering the Biosecurity Modernization and Innovation Act of 2026 to strengthen oversight of synthetic genetic material. Supporters of the letter say the unified stance from AI leaders and biotech firms sends a bipartisan message that stronger biosecurity measures are both necessary and achievable.
Automated bots now account for the majority of global web traffic, with new Cloudflare radar data showing 57.4% of HTML requests coming from bots and just 42.6% from human visitors in the seven days ending June 5th, 2026
Within that bot majority, 50.6% of total traffic is driven by AI training crawlers, while only 10.7% comes from search-related bots, underscoring a sharp imbalance in how content is being consumed. Analysts say the shift happened faster than expected, marking a structural change in the economics of publishing online. Unlike search crawlers that can theoretically send users back to a website, training bots extract content to improve AI models without generating referral traffic, attribution or revenue. Researchers have found that these AI crawlers issue high-volume, parallel requests and scan low-traffic pages, increasing hosting costs while bypassing the caching systems publishers rely on to manage load. Marketing commentator Cyrus S described the situation as a tax on websites, arguing that AI firms monetise extracted data while site owners foot the infrastructure bill. At the same time, a separate debate is intensifying over whether being mentioned in AI-generated answers carries any measurable commercial value. Digital analytics professional Himanshu Sharma contends that if AI mentions truly offset lost search clicks, companies would see stable leads and sales despite falling traffic, which many are not reporting. Industry data shows click-through rates dropping sharply when AI summaries appear in search results, with top positions seeing reductions of more than 50 percent in some studies. Together, the rise of training heavy bot traffic and declining organic clicks is forcing publishers and marketers to reassess whether AI visibility alone can sustain the businesses that create the web's content.

**Carly Wilson** (3:47)
Google appears to be developing a dedicated troubleshooting mode for its Gemini AI, expanding the tool's growing list of capabilities. Reports from users on X indicate the new mode is visible in the Gemini model picker menu for some accounts. When selected, it reportedly provides troubleshooting guidance through a mix of text responses and interactive widgets. Several users claim they can already access the feature, though others say it is not yet available on their accounts.
This suggests the rollout may be accidental, part of a limited test or subject to reversal. Examples shared online show the mode helping users diagnose issues such as a car that will not start. In those cases, Gemini presents a likely cause and offers selectable options so users can specify symptoms. Based on those inputs, the AI then refines its advice and suggests next steps. Discussions on Reddit suggest the mode operates with a lower response temperature, aiming for accuracy and focus rather than conversational flare. While Google has not confirmed the feature, observers say it could become a practical addition to Gemini's toolkit.
OpenAI is rolling out a major upgrade to ChatGPT's memory system, aiming to make the chatbot better at remembering and using personal context. The updated architecture enhances how ChatGPT recalls, synthesizes, and carries forward information across conversations. It builds on the company's existing dreaming process, which allowed the AI to gather useful background details from previous chats. Now users will see a readable summary of what ChatGPT knows about them and can update or refine that information directly. They will also be able to set preferences for which topics the AI should remember and when to reference them. The upgrade addresses a key limitation in earlier versions, where ChatGPT might retain outdated assumptions, such as thinking a user was still on a trip after it had already ended.