Cognitive Warmup. Alphabet and Google CEO Sundar Pichai was at Stanford University to deliver the 2026 commencement speech. If I look beyond the excitement of the speech and the official blog post accompanying the entire speech, Pichai faced a massive student walkout and was heavily criticized during Stanford University’s 135th convocation ceremony at Stanford Stadium on June 14. Reason? Ai.
To some extent, Pichai’s complete avoidance of mentioning AI during his address is perhaps the biggest sign yet – audiences are no longer buying the pitch. Secondly, Google’s involvement in Project Nimbus has not had as many fans as the tech giant had hoped. Project Nimbus is a $1.2 billion cloud computing and artificial intelligence contract with the Israeli government and military that is believed to enable surveillance and military operations.
“I know today is all about giving you advice. But people are also giving me a lot of advice about what to say. In fact, it’s the same advice, and it’s about what not to say. People thought it would be really hard for me; after all, it’s the last two letters of my last name,” Pichai said in an effort to diffuse the widespread tension that is evident.
First on Wired Knowledge
circular tattletale
Something interesting happened a few days after Anthropic released the Fable 5 and Mythos 5 models last week, notably the former was said to be “state-of-the-art on almost all tested benchmarks of AI capability”, and the latter was said to be “the same underlying model as the Fable 5, but with security measures removed in some areas.” That’s when the party really started.
Amazon researchers have reportedly found a way to bypass security measures in Anthropic’s Fable 5 models. There are indications that Amazon has findings that indicate the model could be used to obtain information related to cyber attacks. Amazon CEO Andy Jassy reportedly called senior people at the White House, raising concerns about Anthropic’s new models. Shortly thereafter, Anthropic received a government directive to disable access to the Fable 5 and Mythos 5 – the US administration’s assessment appears to be that both of these models have crossed some performance threshold that no one else has so far, and therefore caution is necessary.
Of course, Anthropic is not pleased at all. By the way, Amazon has invested about $8 billion in Anthropic and is its largest investor. Second, here’s the scenario- Amazon got a jailbreak called the White House, shut down a competitor, and protected its own products. The question I’m asking here is, have the same standards been applied to every AI model that has claimed cutting-edge capabilities so far (and there have been many for funding and enthusiasm continuity) and have all Frontier models that are Fable 5 and Mythos 5 competitors been tested using the same methodology?
Anthropic, noting immediate compliance with the export control directive, states that “We disagree that the discovery of a narrow potential jailbreak should be reason to recall deployed commercial models for hundreds of millions of people. If this standard were implemented industry-wide, we believe it would essentially halt the deployment of all new models for all frontier model providers.”
Vibes and hallucinations
In October last year, KPMG, one of the “big four”, published a report titled Total Experience: Redefining Excellence in the Age of Agentic AIWho writes about how companies are using AI to solve customer demands and make the world a better place. Turns out, he was full of hallucinations. GPTZero, maker of the popular AI content detection tool, found inaccuracies and fake footnotes scattered throughout the report.
GPTZero says that only five of 45 citations in the paper point accurately to actual sources. In at least 28 citations, titles were paraphrased or fake components were added to actual sources, 12 of which were presented intentionally vaguely to make it even more difficult to ensure accuracy. GPTZero researchers also called the creation of fake contexts by AI models as “vibe inference.”
The researchers also found that almost half of the claims in the paper were fake or misrepresented, and “likely the result of the AI research tool being overly compliant with requests to find examples of ‘agent AI’ in the wild”.
As an Emirates loyalist, I know this well. KPMG claimed that Emirates has launched a mobile chatbot called Sara, powered by AI, which can chat with passengers and change their flights. Turns out, Sara, a mobile assistant that was released on the Emirates app and website in 2023, is not an AI-powered chatbot as claimed, and cannot change bookings for passengers.
If I have zero trust in anything “AI”, don’t hold it against me.
my eyes fell on you
If you’ve always wondered why Meta has historically been lenient with its handling of user data. Turns out, this trend runs even deeper within the organization. I read a report a few days ago that suggested that Meta’s internal project called the Model Capability Initiative (MCI) tracks employees on their work devices – that is, every mouse click and keystroke – in an effort to obtain data to improve Meta’s AI models. Turns out, one very slim silver lining in this is that Meta will allow employees to “pause” this tracking for up to 30 minutes when they need to “check on something personal.”
Recently, audio from an all-employee meeting was leaked in which the CEO chatbot defended the program, saying that AI could learn quickly by “watching really smart people do things”, and that “the average intelligence of the people working at this company is significantly higher than other data sets sourced”. A glimpse into the dystopian world that TV shows and movies love to scare us with? They are reality now.