from the real-real-bad dept

As we near the halfway point in the second Trump presidential term, there’s something that is worth remembering: Donald Trump, like most nasty viruses, is a temporary condition. Trumpism may not be, though I have my doubts as to how long a cult of personality can survive without that specific personality leading the cult. But Donald Trump as president will come to an end in the not too distant future.

The millions and millions of people who have been negatively impacted by him and by those who have decided to bow at his cultish altar, are not temporary. They are not going to go away. And they will remember the actions of many during this time.

And I imagine the American Diabetes Association, and specifically those currently leading it, will be in the memories of its members and many others for a long, long time. It’s been nearly a week since the ADA had five diabetes scientists, including its own former president, involuntarily removed from outside the ADA’s annual conference by police. Their crime? Distributing a copy of an editorial from the April edition of the ADA’s own journal.

The scientists were distributing the editorial outside the conference’s opening speech, which was originally scheduled to be given by Jay Bhattacharya, head of the National Institutes of Health under Trump. Bhattacharya canceled at the last minute, and senior NIH official Rick Woychik took his place.

Within minutes of beginning to hand out the editorial, police reportedly escorted the scientists out of the conference, which was held in New Orleans. The police reportedly shoved at least one scientist, took all of their conference badges, and threatened to arrest them if they tried to return. Louisiana State Police later told media that they acted at the request of the ADA. The ADA subsequently barred the five scientists from the rest of the conference.

Advertisement

The editorial just so happened to be very critical of the Trump administration and RFK Jr.’s funding at NIH and other health agencies and groups. It’s quite obvious that the ADA feared repercussions from the Trump administration if it wouldn’t allow these scientists to hand the article out while members of the administration were speaking and tried to use the police to silence them. And then, when this whole thing went viral, the ADA offered up justifications for its actions. Justifications that kept changing, as it turns out.

In an email to ADA members Saturday, the association said the scientists were removed because they didn’t have prior approval to distribute material at the conference and that it was “not because of the viewpoints expressed in those materials,” according to reporting from Science.

In a statement Sunday, the organization, which is a nonprofit, said it removed the scientists because it was complying with federal regulations for 501(c)(3) nonprofits, which requires “maintaining a strictly nonpartisan environment at all organizational events and functions while engaging across party affiliations to advance our mission.” However, the federal regulations do not restrict leaders of organizations from sharing political views in a personal capacity or from speaking on important public policy issues.

And from there, the Streisand Effect took over. The editorial, which you can find right here, went somewhat viral itself, getting a ton more attention than it had to date. But the real backlash came from the public and from within the medical community itself. There have been resignations in protest of the ADA’s actions. An open letter to the ADA signed by 40 members was written to torch leadership’s actions and treatment of the scientists at the conference. Another open letter was also written, likewise demanding an apology.

And, finally, the ADA did in fact apologize days later.

In the video Wednesday, ADA CEO Charles Henderson personally apologized to the five scientists, including Aaron Kelly, pediatrics professor at the University of Minnesota; Justin Ryder of Northwestern University; and Irl Hirsch, also of the University of Washington, in addition to Kahn and Schatz.

“What transpired is not reflective of who I am, the values I hold, or the way I was raised,” Henderson said. “I will work hard to bring our community back together to build on the progress we have collectively made for those affected by diabetes.”

In addition to apologizing to the five ejected scientists, Henderson apologized to the community as a whole, saying that the ADA would commission a “thorough independent review of the events that occurred as well as the policies, procedures, and decision-making process that guided our actions.”

Yeah, no, not good enough. The fish stinks from the head down, as the saying goes, and there have been days worth of attempts to make this stupidity anyone’s fault but leadership at the ADA. This was a clear attempt to lick the Trump administration’s boots, at the very moment when clear leadership from medical groups is so sorely needed, and that’s a bell that cannot be un-rung.

Henderson needs to go. And I have little doubt that he will before too long. Trump and RFK Jr. will eventually be gone, as well.

But we won’t forget how groups like the ADA, and the people leading them, acted during this time.

Filed Under: free speech, jay bhattacharya, nih, protests, streisand effect

Companies: american diabetes association

There’s still the possibility of that Trump pardon.

From connected factory floors to automation and digital twins, ‘Industry 4.0’ refers to the future of manufacturing.

Ireland’s diverse manufacturing industry makes it the choice European location for a growing number of industrial and technology companies, according to IDA Ireland.

It does especially well in some sectors – medtech, automotives, aerospace and chemicals being a few examples – with its capacity only enhanced by a steady supply of skilled talent emerging from third-level institutions.

Meanwhile, a strong talent pipeline supported by grants and commercialisation support for research and innovation also allows Ireland to maintain its appeal as a global manufacturer.

From connected factory floors and industrial IoT to AI-powered automation and digital twins, ‘Industry 4.0’ is a phrase used to refer to the future of manufacturing. Here are seven Irish start-ups innovating on the factory floor.

Forge Robotics

This Galway start-up is a part of the prestigious Y Combinator accelerator programme behind the likes of Airbnb, Stripe and Uber.

Founded by CEO Eoin Cobbe and CTO Robert Cormican, Forge Robotics wants to tackle the rising threat of skilled welder shortages in manufacturing using intelligent automation.

The company makes an AI-powered intelligence layer that improves the welding capabilities of industrial robots. Its system allows robots to scan a part, interpret its geometry and execute welds even when the set-up is imperfect.

Gemell Technology

Headquartered in Dublin’s Dogpatch Labs, this Enterprise Ireland (EI) ‘High Potential Start-Up’ builds 3D visualisation and sustainability software for textile manufacturers and designers.

Gemell Technology aims for its technology to significantly reduce unnecessary fabric samples from ending up in landfills.

The company can generate photo-like digital models of yarns and fabrics, which manufacturers can tweak instead of ordering physical fabric samples. These 3D renders are generated with fabric textures originating from individual fibres.

Gemell claims that manufacturers that use its technology reduce unnecessary fabric samples and waste by 70pc, while getting products to market 11 weeks faster.

The company has offices in Dublin and London, and came in as the first runner-up at last year’s All-Island Circular Venture Awards – a competition that recognises late-stage start-ups across the island showcasing circular value propositions.

InnaLabs

Last year, this Dublin-based space-tech secured its second contract to advance Earth’s planetary defence with the European Space Agency (ESA)’s Ramses mission.

The Rapid Apophis Mission for Space Safety – or Ramses – will rendezvous with asteroid 99942 Apophis, accompanying it during its extremely close but safe flyby of Earth in 2029.

InnaLabs’ gyroscope navigation system will be helping the ESA, the Italian aerospace engineering company OHB and the Spanish tech company GMV in the space mission scheduled for launch in April 2028.

The company’s technology solves complex navigation, stabilisation and guidance challenges within space, aerospace, land and marine markets.

Oscil

Oscil provides data analytics support for the pharmaceutical and dairy sector. The company’s founder Dr Patrick Cronin won the EI Big Ideas award last year after contesting in a pitching battle against other pre-spin-out ventures emerging from EI’s Commercialisation Fund.

Speaking to SiliconRepublic.com, Cronin said that the “current rise in weight loss drugs [and] GLP-1s are driving huge investment in the protein market”, and that “Oscil can unlock a lot of capacity and quality in spray driers through edge sensing and machine learning to provide real-time process control”.

The company said it is seeking early adopters in the spray drying industry to improve production capacity and product quality.

Smartfactory

This 2016-founded start-up captures, analyses and visualises performance indicators from the manufacturing, logistics and utility sectors using Industry 4.0 technology.

The company’s SaaS solutions, built on technology from hardware partners Siemens and Banner, help cut down on manufacturing downtime by identifying hidden losses in the production process.

It is based out of Nexus Innovation Centre at the University of Limerick.

Ubotica

This Dublin-based space-tech is a frequent collaborator with NASA and the ESA. Last month, it announced a partnership with Texas’s Novi Space to deliver real-time intelligence from the Earth’s orbit.

The collaboration enables Earth observation data to be processed directly on satellites instead of it needing to be transferred to Earth for analysis.

Ubotica is deploying its AI platform, which processes input data within 90 seconds, for the space mission. According to the company, in a single test observation of a Singapore port, the platform processed hundreds of vessels and detected those operating ‘dark’ in under two minutes.

The company has deployed its AI capabilities on numerous missions, including its own CogniSAT-6 satellite.

WrxFlo

Founded in 2019 by former Dell manufacturing leaders Tim Crowe, Ken Sheehan and Jennifer Kelly, WrxFlo is a SaaS platform tailored specifically for manufacturing and logistics operations.

The company claims its platform acts as a “digital co-worker” across operations and the vertical line of command by connecting data from across the factory, warehouse and supply chain, eliminating non-value-added tasks and surfacing ‘red’ indicators before they become costly problems.

“We built WrxFlo from first-hand experience of running complex manufacturing and supply chain operations,” said Crowe, the company’s CEO.

“WrxFlo enables industrial manufacturers and logistics operators turn complex, paper or Excel-based processes into streamlined, data-driven systems that reduce cost and improve efficiency.”

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

A former Fitbit executive is betting that the next big wearable might not be a smartwatch at all.

Stacy Salvi, who spent eight years at Fitbit before and after Google’s acquisition of the company, has launched The Gem. It is a smart necklace designed to help women track their UV exposure throughout the day.

Rather than counting steps or monitoring workouts, The Gem focuses on something many wearables overlook: how much sunlight your skin is actually getting. The pendant-style device continuously measures both UVA and UVB exposure and sends the data to a companion app. As a result, the app builds a personalised skin profile based on your habits and environment.

The idea is to provide more useful guidance than a simple weather forecast. Depending on your exposure levels, the app can recommend sunscreen reminders, highlight safer times to spend outdoors and help users understand their personal UV limits.

OpenText’s €105m investment marks the single largest backing by a company with a Canadian HQ.

Canadian enterprise data management company OpenText is creating 400 highly-skilled jobs in Cork and Galway over the next three years through a planned investment of €105m.

The company helps organisations protect, govern and utilise their data for agentic AI, cybersecurity and sovereign cloud. Irish-based developers and researchers will design, deploy, secure and operate these AI and cloud capabilities for the European, Middle East and African (EMEA) markets.

This is the single largest investment into Ireland by a technology company headquartered in Canada and is supported by IDA Ireland.

The major investment is expected to advance the company’s service capabilities across the EMEA industries and public sectors. It is also expected to increase regional capacity for organisations operating in highly regulated and mission-critical environments which require greater control over data governance.

“This investment expands our EMEA R&D and operations capacity to deliver the trusted AI, cybersecurity, and cloud capabilities our clients already rely on globally, while giving European organisations greater regional support and flexibility across the cloud environments of their choice,” said Shannon Bell, the executive vice-president, chief digital officer and chief information officer of OpenText.

Speaking of the announcement, Michael Lohan, the CEO of IDA Ireland said: “This investment will strengthen Ireland’s leadership in AI and transformational technology and IDA Ireland looks forward to continuing to work closely in partnership with OpenText as it grows its business in Ireland and deepens its European presence.”

OpenText said that it intends to explore opportunities for university and research collaboration in Ireland as part of its long-term innovation and talent strategy. This includes partnerships with third-level institutions focused on AI, cybersecurity and secure digital operations.

Laying down its R&D plans, OpenText said that its research efforts will advance how AI agents are orchestrated and governed including multi-agent collaboration, system boundary enforcement and knowledge sharing across sovereign zones.

In data sovereignty, it is developing continuous compliance mechanisms that give organisations verifiable control over where data lives and how it is governed, while its cybersecurity research will focus on threat detection and response.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

OpenAI is under investigation by a coalition of state attorneys general, according to the Wall Street Journal. On Friday, June 12, the company received a subpoena seeking information and documents related to its activities and impact on users. The Journal said it viewed the subpoena sent by New York’s attorney general. 

Based on what the publication saw, the AGs are asking for documentation about the company’s advertising, user engagement and retention, as well as its handling of its users’ data and health information. They also want to know about the company’s activities related to minor and senior users, its deep learning models, its policies and its models’ sycophancy. 

“AI is a new and powerful technology, and we work every day to safely bring its benefits to people in a responsible way,” an OpenAI spokesperson said in a statement to the Journal. “We take the concerns raised by state attorneys general seriously and intend to engage constructively with their offices.”

It’s unclear what prompted the investigation, but tech companies developing AI products have been under scrutiny by state AGs for quite a while now. Last year, a group of 44 state AGs sent a letter to Meta, Google, Apple, Microsoft, OpenAI, Anthropic, Perplexity AI and XAI, asking them to protect children from being exposed to inappropriate and potentially harmful chatbot interactions. In April, Florida Attorney General James Ulthmeier opened a criminal investigation into OpenAI, because the suspect in the 2025 Florida State University mass shooting reportedly used ChatGPT.

More recently, another parent filed a wrongful death lawsuit against OpenAI, accusing the company of not implementing enough safeguards to protect users from taking their own life. The lawsuit claimed that the plaintiff’s daughter who died by suicide discussed her suicidal thoughts and plans with the chatbot in the months leading up to her death. However, the company didn’t alert the family or authorities. OpenAI was named as a defendant in the first ever wrongful death lawsuit linked to a chatbot, as well.

Just a few days ago, OpenAI filed paperwork with the Securities and Exchange Commission to go public. It hasn’t decided on timing and pricing yet.

OpenAI is forcing Mac users to update ChatGPT and other desktop apps, after a supply chain attack exposed signing certificates that Apple’s security systems use to verify trusted software.

The company disclosed the incident on May 13 and confirmed malware linked to the “Mini Shai-Hulud” attack infected two employee devices through the TanStack npm ecosystem. Investigators identified unauthorized access activity in a limited set of internal source code repositories connected to those employees.

OpenAI rotated its signing certificates and re-signed affected apps to prevent potential misuse of the exposed credentials. The company found no evidence that customer data, production systems, or intellectual property were compromised during the incident.

Apple’s macOS security protections will block apps signed with the older certificates after June 12, which makes the update mandatory for affected Mac users.

OpenAI confirmed the affected repositories included signing certificates used for applications across macOS, iOS, Windows, and Android. The company blocked future notarization attempts tied to the older credentials instead of revoking the certificates immediately and risking broken software installations for existing users.

Mac users must install updated versions before June 12. After that date, Apple’s security protections will stop trusting apps signed with the previous certificates.

Why macOS users need to update

Code-signing certificates help macOS verify that software comes from a legitimate developer. Apple’s Gatekeeper and notarization systems use those certificates to determine whether apps should be trusted, launched, or blocked.

Investigators found no evidence that exposed certificates were used to sign malicious software or distribute malware to users. OpenAI reviewed prior notarizations for signs of unauthorized activity and said it found no evidence of misuse.

Older versions of ChatGPT Desktop, Codex App, Codex CLI, and Atlas signed with the previous certificates may stop functioning or receiving updates after June 12. ChatGPT Desktop 1.2026.125, Codex App 26.506.31421, Codex CLI 0.130.0, and Atlas 1.2026.119.1 are the affected releases.

Supply chain attacks are becoming harder to contain

Modern apps rely on vast networks of open-source libraries, package managers, and automated development systems that can spread compromised code widely. A malicious dependency can traverse multiple organizations before developers detect the malware in the software chain.

Apple’s macOS security protections will block apps signed with the older certificates after June 12

The attack hit during an active rollout of new supply chain security protections across OpenAI’s development systems. Those protections included stricter package provenance checks, stronger CI/CD credential controls, and package-manager safeguards like minimumReleaseAge policies.

The two affected employee devices hadn’t yet received the updated protections when the malware reached the systems. OpenAI said the incident accelerated deployment of additional safeguards designed to reduce the impact of future supply chain attacks.

How Mac users can stay safe

OpenAI told users to install updated apps only through official websites or built-in update systems. The company also warned users to avoid installers distributed through ads, third-party download sites, email links, or unsolicited messages.

Mac users should verify they are running the latest versions of ChatGPT, Codex, and related OpenAI apps before June 12. Users who downloaded OpenAI software from unofficial sources should delete those apps and reinstall clean versions directly from OpenAI.

Most enterprise RAG pipelines start the same way: a text parser converts web pages and documents into plain text so they can be chunked and indexed for retrieval. That conversion step destroys retrieval signals — and according to new research, it’s responsible for the majority of wrong answers.

A research team from UC Berkeley, Princeton University, EPFL and Databricks published a paper this week introducing PixelRAG, a system that skips that conversion entirely. Instead of parsing pages into text, PixelRAG renders them as screenshots, indexes those images and feeds retrieved tiles directly to a vision-language model reader. Tested across 30 million screenshot tiles covering all of Wikipedia, it outperforms text-based RAG across six benchmarks, improving accuracy by up to 18.1% over text-based baselines.

Parsers are the wrong place to look for fixes, according to the research team.

“Improving parsers is an endless process because every website requires special handling,” Yichuan Wang, lead author and UC Berkeley doctorate student, told VentureBeat.  “Our goal was to explore whether recent advances in VLMs make it possible to bypass that entire problem and build a retrieval system that works across websites without site-specific engineering.”

HTML parsers destroy the retrieval signals that enterprise RAG depends on

The goal of the researchers was to develop a clean end-to-end architecture.

“Modern web RAG pipelines often involve rendering, parsing, cleaning, chunking, and many other handcrafted stages,” Wang said. “Every stage introduces potential cascade errors and abstractions that move us further away from the original webpage. We were interested in whether we could eliminate most of that complexity and operate directly on the rendered page.”

Wang also noted that parsing inevitably loses information. Images, visual hierarchy, typography, emphasis (e.g., bold text), tables, and layout are either discarded or converted into imperfect textual approximations. 

“No matter how good a parser becomes, some information is fundamentally lost during the conversion,” he said.

The research identifies three ways text-based RAG loses the answer before it reaches the reader. All three were measured on SimpleQA, a standard benchmark of 1,000 factual Wikipedia questions:

• Parser loss (36.6% of failures). HTML-to-text conversion destroys structured content so completely that no text chunk in the corpus contains the answer.

• Rank loss (55.2% of failures). The answer exists in the corpus but gets outranked by keyword-dense infoboxes that land at rank 1 for 75.9% of queries, pushing answer-bearing paragraphs to rank 20 or lower.

• Reader loss (8.2% of failures). The correct content reaches the reader but flattened structure causes misattribution.

How PixelRAG works 

Unlike a standard LLM that reads only text, a vision-language model takes images as input alongside text, meaning it can read a rendered web page the way a human does, with layout and structure intact. “For many structured information extraction tasks, we believe modern VLMs have an inherent advantage because they can reason jointly over both content and layout rather than relying on a flattened text representation,” Wang said.

PixelRAG is built around that principle, replacing the text parsing pipeline with a four-stage system that operates entirely on rendered screenshots.

• Rendering. Pages are rendered using Playwright, a browser automation library, at a fixed 875-pixel viewport and sliced into 1024-pixel-tall tiles. Wikipedia’s 7 million articles produce roughly 30 million tiles. Assets are cached locally and rendered entirely offline.

• Indexing. Each tile is encoded as a single 2048-dimensional vector using Qwen3-VL-Embedding-2B and stored in a FAISS approximate nearest-neighbor index. The full index runs to approximately 120 GB in fp16 and supports incremental updates without full re-indexing.

• Training. The retrieval model is fine-tuned on synthetic contrastive data generated from the datastore, using dynamic hard-negative mining to filter false negatives. LoRA, a lightweight fine-tuning method that updates a small fraction of model weights, is applied to both the language model backbone and the visual encoder. Training on approximately 40,000 pairs completes in under three hours on a single H100.

• Storage. Raw screenshot tiles for Wikipedia require 5.6 TB, but a render-on-demand approach eliminates persistent storage: embed all tiles, delete the screenshots and re-render pages on demand at query time. The vector index requires approximately 120 GB.

Six benchmarks, 10x agent token savings and one unsolved problem

Researchers tested PixelRAG across six benchmarks spanning factual Wikipedia QA, table-based queries, multimodal QA and live news retrieval. They said it outperformed text-based RAG on all six, including on tasks where questions are answerable from text alone. On SimpleQA it reaches 78.8% accuracy versus 71.6% for the strongest text parser, widening to 48.8% versus 42.5% on structured table queries. Teams need Qwen3-VL-4B class models or above to see the benefit. Smaller models trail text retrieval by more than 12.5 percentage points.

The agent cost advantage is the strongest near-term case for PixelRAG. In benchmark testing, an AI agent using PixelRAG as its search backend ran on 3.6 million prompt tokens versus 37.5 million for text retrieval, at 2 to 4 times lower cost than alternatives including Google, while achieving higher accuracy. Image compression can cut that token budget by a further third.

Visual chunking is the main unsolved problem. Text-based RAG systems have spent years refining how to split documents into meaningful retrieval units based on topic, section or semantic content. PixelRAG currently has no equivalent: it slices pages by fixed pixel height, meaning a table or paragraph can get cut in half mid-tile with no awareness of content boundaries. 

“The text retrieval community has spent years studying chunking strategies, while visual retrieval has received much less attention,” Wang said. “We think this is an important area for future research.”

What this means for enterprises

The retrieval quality problem PixelRAG addresses reflects a broader market shift already underway. VB Pulse Q1 2026 data from qualified enterprise respondents found intent to adopt hybrid retrieval tripling from 10.3% in January to 33.3% in March, the fastest-growing strategic position in the dataset. PixelRAG’s own authors point to hybrid deployment as the most practical near-term path — layering visual retrieval on top of existing text systems rather than replacing them.

For teams already running RAG pipelines, the path to those savings is more straightforward than a ground-up rebuild.

“A practical path is to use PixelRAG as an enhancement layer alongside existing text retrieval systems,” Wang said. “Hybrid retrieval that combines both text and visual search is straightforward and is likely how many production deployments would evolve.”

Maine has taken its public data breach reporting portal offline after fraudulent breach disclosures were published on the state’s website, prompting a review of procedures to prevent abuse in the future.

Yesterday, BleepingComputer reported that fake data breach disclosures had been submitted to Maine’s official breach notification portal impersonating Discord and the multiplayer social virtual reality platform VRChat.

At the time, VRChat told BleepingComputer the filing was fraudulent and had been submitted using the name of a fictitious employee.

In a statement published Friday, the Maine Attorney General’s Office acknowledged that data breach “hoaxes” were submitted through the state’s reporting system.

“The Office of the Maine Attorney General has been made aware of an apparent abuse of our data breach reporting system,” the statement reads.

“After conversations with VRChat, one of two affected companies, it has become clear that the reported data breaches were hoaxes submitted by an unknown entity unrelated to either company. These false reports have been removed from the database. We have no knowledge of any recent legitimate data breach reports from either VRChat or Discord.”

The Attorney General’s Office says it has now temporarily disabled public access to the breach notification database while it reviews reporting procedures to reduce similar abuse in the future.

Prior to the shutdown, submitted breach notices were automatically published to the public database.

“We don’t have any independent knowledge of the breaches, the submitting entity fills out the information and it goes directly onto the site. We will review the one you’ve flagged, thank you,” Maine Attorney General’s Office told BleepingComputer.

The notice states that companies can continue to submit breach notifications through the reporting service, but members of the public seeking copies of disclosures must now contact the Attorney General’s Office directly.

Maine’s data breach portal is commonly used by journalists, researchers, and threat intelligence firms to monitor newly disclosed security incidents and determine whether organizations are reporting cyberattacks or data breaches affecting consumers.

The incident demonstrates how automatically published breach disclosures can be abused to spread misinformation and damage a company’s reputation.

The fraudulent VRChat filing claimed the company suffered a data breach impacting over 2.4 million people and included a fabricated employee contact name in the disclosure.

After BleepingComputer contacted VRChat about the filing, the company confirmed the disclosure was fake and stated it had not submitted the notice to Maine authorities.

BleepingComputer also contacted Discord about the fraudulent notice submitted to the site but did not receive a response.

It is unclear how many additional fraudulent breach notices may have been submitted through the portal before the state suspended public access to the database.

Security teams log 54% of successful attacks and alert on just 14%. The rest move through your environment unseen.

The Picus whitepaper shows how breach and attack simulation tests your SIEM and EDR rules so threats stop slipping by detection.

Advertisement

Get the whitepaper