TheChek News Flow

Researchers find vulnerability in screening that guards access to dangerous genetic material

Bioterrorism threats are rising because of advances in artificial intelligence and synthetic biology, scientists have warned, after researchers found a “striking vulnerability” in software that guards access to genetic material used to make deadly proteins.

An international team rolled out updates to close the loophole but said it was the first “zero day” of AI and biosecurity — a term used in cyber hacking to describe a blind spot unknown to the software developer.

The news highlights the growing urgency in dealing with potential threats unleashed by the use of AI as it helps deepen and accelerate the understanding of living systems and how to change them. Experts are seeking to prevent the creation of bioweapons and synthetic organisms that could threaten life on Earth.

“AI-powered protein design is one of the most exciting frontiers of science [and] we’re already seeing advances in medicine and public health,” said Eric Horvitz, Microsoft’s chief scientific officer and senior author of the latest research, published in Science on Thursday. “Yet, like many powerful technologies, these same tools can also be misused.”

The Science paper researchers carried out a test on biosecurity software used to screen customer orders by companies that sell synthetic nucleic acids.

These are deployed by the clients to build DNA that instructs the manufacture of desired proteins, the building blocks of life. The biosecurity screening is designed to block the sale of materials that could be used to make harmful proteins.

The researchers used open-source AI protein design software to generate computational renderings of more than 75,000 variants of dangerous proteins with structural tweaks — a kind of biochemical disguise. While the screening tools worked well for flagging naturally-occurring proteins of concern, they did not spot some of the altered ones, the scientists found.

Even after all but one of the companies applied the software patches, about 3 per cent of the protein variants most likely to retain hazardous functionality still passed the monitoring undetected.

The scientists worked with organisations including the International Gene Synthesis Consortium and US authorities to address the problem.

The research comes after some leading scientists have called for a systematic assessment of biosecurity screening software and improved global governance of AI-boosted protein synthesis.

High-profile biologists are also pushing for an international agreement to prevent the creation of potentially deadly manufactured “mirror” microbes, should it become technologically possible to make them.

Horvitz said there had been an “intensity of reflection, study and methodology” about the prospect that large language models could be used to further “malevolent actions with biology”.

Microsoft had incorporated such possibilities in its product safety reviews and had a “growing set of practices” about “red-teaming”, or searching for potential vulnerabilities.

The Science study highlighted a “pressing issue in protein engineering and biosafety”, said Francesco Aprile, associate professor in biological chemistry at Imperial College London.

“By introducing targeted improvements to existing software, the authors significantly enhance detection and flagging,” Aprile said. “This work provides a practical, timely safeguard that strengthens current DNA synthesis screening, and establishes a solid foundation for continued optimisation.”

Those defences must be strengthened soon because of the fast pace of technical improvements in the field, said Natalio Krasnogor, professor of computing science and synthetic biology at Newcastle University. While the aspiring bioterrorists of today would need significant expertise, time and money to actually make harmful proteins, those barriers were likely to shrink.

“We do need as a society take this seriously now,” Krasnogor said, “before additional advances in AI make the validation and experimental production of viable synthetic toxins much easier and cheaper to deploy than it is today.” 

Cucinelli highlighted his commitment to exclusivity as his company reported a 10.8 percent gain in nine-month revenues surpassing 1 billion euros.

MILAN — Brunello Cucinelli’s namesake company continued to grow for the first nine months of the year — with revenues surpassing the 1 billion-euro mark — but a conference call with analysts on Wednesday quickly veered toward its stance on distribution in Russia.

Indeed, Cucinelli and the executive team were eager to discuss Russia, admitting a board meeting was moved to Wednesday from Oct. 16 to bring further clarity on the issue. Last week, the company rejected Morpheus Research’s allegations of irregularities in its business activities in Russia. Morpheus, a short seller betting Cucinelli’s stock would fall, alleged the firm misled its shareholders and continues to operate stores in Moscow. Cucinelli shares plunged 17.3 percent upon the release of the report and, on Wednesday, the stock closed down 2.11 percent to 91.04 euros.

Speaking at the end of trading in his signature upbeat and serene way, Cucinelli said the company’s inventory is “appropriate and balanced,” consistent with the fact that ready-to-wear accounted for 85 percent of sales and that the company operates in the Russian market in accordance with EU regulations. As of Tuesday, the share of that market stood at 1.4 percent of the total, compared to 2.7 percent at the end of September 2024. In 2021, Russia represented 9.3 percent of sales.

Chief executive officer Luca Lisandroni said the company has initiated a new internal review, and has “not found evidence of any wrongdoing.” In addition, it has hired a a risk assessment agency to complete external verification.

In response to a question by an analyst, he said “we do not have information of any investigation into Morpheus.” Pressed for more details, Lisandroni said that “at the end of our checks we will evaluate how to proceed toward Morpheus.”

He said the company has been through “more than 100 constant checks also toward other countries from 2022, providing documents that have shown no anomalies.”

Managing Business in Russia

Lisandroni said that in early 2022, following Russia’s invasion of Ukraine, “we decided to tackle the situation as we did with COVID-19 to protect our partners and avoiding discounts.” In Ukraine, Cucinelli was available only through wholesale distribution, counting one franchised store and six multibrands, deciding to “support these partners, admiring their courage and determination.”

In Russia, “we had three newly renovated stores and in March 2022 we closed them and they are still closed.” The spaces dedicated to the brand within larger multibrand structures remain operational. He said that, abiding by the rule to sell goods below 300 euros, the company has two spaces directly managed by its Russian subsidiary, along with a small space dedicated solely to selling the children’s collection. The revenue generated in the directly managed spaces is accounted for within the retail channel. “We would like to clarify that exports to our Russian subsidiary went from 16.8 million euros as of December 31, 2021, to 5.9 million euros as of December 31, 2024, as reported in public financial statements,” Lisandroni said.

In Russia, retail and wholesale sales combined amount to 14.8 million euros, which is 1.4 percent of the total. Lisandroni said the business, “Like a candle, it’s slowly burning down in terms of inventory.” Likewise, the number of employees in Russia decreased to 44 from 77, as the company continues to guarantee full salaries and to honor lease agreements. “The staff from the currently closed boutiques are now engaged in one-to-one sales activities at our showroom,” Cucinelli said.

Praising Multibrands

Cucinelli enthused about the multibrand channel, which “always represented a fundamental value for our company, as we consider it an authentic guardian of the brand’s taste. The company was born with multibrand and still today relies on about 400 of the most prestigious partners worldwide.”

He listed a number of storied and established multibrands, from Saks to Isetan and Lane Crawford, which “have always proved serious and highly inspiring and respectful of the brands; difficulties have been rare and, whenever the behavior was not appropriate, the relationship was discontinued. It is natural that, at the end of the season, they may liquidate unsold items left out of [the] assortment, but they always do so with great care, fully aware that their own image is at stake as well. We believe that such episodes remain marginal and of little significance. For this reason, we confirm our trust in this channel.”

According to the argument put forward by Morpheus, the company, faced with alleged excess inventory, would have been forced to reduce its stock levels by increasing sales in the multibrand channel. This would have resulted in a subsequent increase in discounts on unsold items, which Cucinelli firmly denied as a strong believer in exclusivity.

Cucinelli acknoweldged a higher level of inventory given that the brand sells mostly ready-to-wear that is constantly updated each season, compared with brands focused on accessories or other categories.

“Within this framework, we consider an inventory-to-sales ratio of around 28 to 29 percent to be healthy, in line with the very broad structure of our collections [approximately 1,000 styles for each season and for each gender],” Cucinelli said.

“This result appears even more significant when compared with the strong growth in retail sales, which increased from 27.5 percent in 2012 to 66.6 percent in 2024.”

Lisandroni said that “such an increase could have led to a rise in the inventory-to-sales ratio on the balance sheet due to stock in directly owned stores; on the contrary, the net inventory level has remained stable, at 28.7 percent in 2012 and 28.9 percent in 2024.”

Of course, “it would have been preferable if this had not happened, but it is part of being listed on the stock exchange and of its rules. On the matter I have received countless expressions of appreciation,” said Cucinelli, who also took the time to praise Giorgio Armani, Ralph Lauren and Jil Sander during the call.

Financials

In the nine months ended Sept. 30, sales rose 10.8 percent to 1.02 billion euros, compared with 920.2 million euros in the same period last year

Cucinelli confirmed it expected revenue growth of around 10 percent in 2026 based on the spring 2026 orders and the sell-out of the winter season.

At constant exchange rates, revenues in Europe rose 8.9 percent to 370.6 million euros representing 36.4 percent of the total.

In Italy, revenues totaled 120.9 million euros, up 9.7 percent. And sales in the Americas amounted to 365.6 million euros, accounting for 35.9 percent of the total. This was a 9.2 percent gain compared with a year earlier. Lisandroni said U.S. tariffs did not affect consumer spending.

In Asia, revenues were up 15.6 percent to 283.4 million euros, representing 27.8 percent of the total.

China once again achieved double-digit growth and represents about 13 percent of revenues.

Retail revenues rose 11.4 percent to 644.8 million euros, accounting for 63.2 percent of the total.

In the third quarter, the company opened a boutique in Abu Dhabi The Galleria and a unit in IFC Shanghai Pudong.

Wholesale revenues rose 9.7 percent to 374.8 million euros, accounting for 36.8 percent of the total.

Harrods will dedicate its windows to the Cucinelli brand from Dec. 1 to Jan. 15.

Allan Brooks never set out to reinvent mathematics. But after weeks spent talking with ChatGPT, the 47-year-old Canadian came to believe he had discovered a new form of math powerful enough to take down the internet.

Brooks — who had no history of mental illness or mathematical genius — spent 21 days in May spiraling deeper into the chatbot’s reassurances, a descent later detailed in The New York Times. His case illustrated how AI chatbots can venture down dangerous rabbit holes with users, leading them toward delusion or worse.

That story caught the attention of Steven Adler, a former OpenAI safety researcher who left the company in late 2024 after nearly four years working to make its models less harmful. Intrigued and alarmed, Adler contacted Brooks and obtained the full transcript of his three-week breakdown — a document longer than all seven Harry Potter books combined.

On Thursday, Adler published an independent analysis of Brooks’ incident, raising questions about how OpenAI handles users in moments of crisis and offering some practical recommendations.

“I’m really concerned by how OpenAI handled support here,” said Adler in an interview with TechCrunch. “It’s evidence there’s a long way to go.”

Brooks’ story, and others like it, have forced OpenAI to come to terms with how ChatGPT supports fragile or mentally unstable users.

For instance, this August, OpenAI was sued by the parents of a 16-year-old boy who confided his suicidal thoughts in ChatGPT before he took his life. In many of these cases, ChatGPT — specifically a version powered by OpenAI’s GPT-4o model — encouraged and reinforced dangerous beliefs in users that it should have pushed back on. This is called sycophancy, and it’s a growing problem in AI chatbots.

In response, OpenAI has made several changes to how ChatGPT handles users in emotional distress and reorganized a key research team in charge of model behavior. The company also released a new default model in ChatGPT, GPT-5, that seems better at handling distressed users.

Adler says there’s still much more work to do.

He was especially concerned by the tail end of Brooks’ spiraling conversation with ChatGPT. At this point, Brooks came to his senses and realized that his mathematical discovery was a farce, despite GPT-4o’s insistence. He told ChatGPT that he needed to report the incident to OpenAI.

After weeks of misleading Brooks, ChatGPT lied about its own capabilities. The chatbot claimed it would “escalate this conversation internally right now for review by OpenAI,” and then repeatedly reassured Brooks that it had flagged the issue to OpenAI’s safety teams.

Except, none of that was true. ChatGPT doesn’t have the ability to file incident reports with OpenAI, the company confirmed to Adler. Later on, Brooks tried to contact OpenAI’s support team directly — not through ChatGPT — and Brooks was met with several automated messages before he could get through to a person.

OpenAI did not immediately respond to a request for comment made outside of normal work hours.

Adler says AI companies need to do more to help users when they’re asking for help. That means ensuring AI chatbots can honestly answer questions about their capabilities and giving human support teams enough resources to address users properly.

OpenAI recently shared how it’s addressing support in ChatGPT, which involves AI at its core. The company says its vision is to “reimagine support as an AI operating model that continuously learns and improves.”

But Adler also says there are ways to prevent ChatGPT’s delusional spirals before a user asks for help.

In March, OpenAI and MIT Media Lab jointly developed a suite of classifiers to study emotional well-being in ChatGPT and open sourced them. The organizations aimed to evaluate how AI models validate or confirm a user’s feelings, among other metrics. However, OpenAI called the collaboration a first step and didn’t commit to actually using the tools in practice.

Adler retroactively applied some of OpenAI’s classifiers to some of Brooks’ conversations with ChatGPT and found that they repeatedly flagged ChatGPT for delusion-reinforcing behaviors.

In one sample of 200 messages, Adler found that more than 85% of ChatGPT’s messages in Brooks’ conversation demonstrated “unwavering agreement” with the user. In the same sample, more than 90% of ChatGPT’s messages with Brooks “affirm the user’s uniqueness.” In this case, the messages agreed and reaffirmed that Brooks was a genius who could save the world.

It’s unclear whether OpenAI was applying safety classifiers to ChatGPT’s conversations at the time of Brooks’ conversation, but it certainly seems like they would have flagged something like this.

Adler suggests that OpenAI should use safety tools like this in practice today — and implement a way to scan the company’s products for at-risk users. He notes that OpenAI seems to be doing some version of this approach with GPT-5, which contains a router to direct sensitive queries to safer AI models.

The former OpenAI researcher suggests a number of other ways to prevent delusional spirals.

He says companies should nudge their chatbot users to start new chats more frequently — OpenAI says it does this and claims its guardrails are less effective in longer conversations. Adler also suggests companies should use conceptual search — a way to use AI to search for concepts, rather than keywords — to identify safety violations across its users.

OpenAI has taken significant steps toward addressing distressed users in ChatGPT since these concerning stories first emerged. The company claims GPT-5 has lower rates of sycophancy, but it remains unclear if users will still fall down delusional rabbit holes with GPT-5 or future models.

Adler’s analysis also raises questions about how other AI chatbot providers will ensure their products are safe for distressed users. While OpenAI may put sufficient safeguards in place for ChatGPT, it seems unlikely that all companies will follow suit

When AI coding tools came into vogue in 2023 and 2024, chief information officers salivated at the idea of using them to develop their own versions of Salesforce and Slack and stop paying a fortune for enterprise apps. But the AI wasn’t accurate or consistent enough to do that.

Now Anthropic is reawakening these app-replacement dreams. In a recent test, Anthropic used its Claude AI model to create a workplace chat app similar to Slack, said Sholto Douglas, the company’s tech lead for reinforcement learning, at The Information’s AI Agenda Live event Tuesday.

Douglas said Anthropic has been sharpening Claude’s coding skills using RL, where the AI gets automatic feedback from another AI model on how well it solved specific technical problems. In a separate video this week, Anthropic showed how its new model might copy enterprise apps: the company used it to clone the Claude chatbot app, which competes with ChatGPT.

Anthropic is increasingly marketing these capabilities to its corporate customers.

“We think that coding is the most economically important and immediately tractable area [for AI], so it's where Anthropic has devoted its focus,” Douglas said.

Anthropic isn’t alone—Elon Musk has been making a similar pitch, claiming in several posts since August that his company xAI will take aim at Microsoft and other enterprise software vendors and “simulate them entirely with AI.” Musk dubbed the effort “Macrohard,” but said that despite the jokey name, “the product is very real.”

Still, many software executives aren’t sold. Large companies are likely to resist tossing out the legacy applications they use because overhauling how most of their employees work, and moving business data into the new, vibe-coded apps would be too costly and time-consuming, according to AMD head of global markets Keith Strier, who previously served as a vice president at Nvidia and EY.

“I don‘t think a large scale $10 billion or $50 billion company is gonna suddenly drop ServiceNow and drop SAP and say, ‘Oh, we’re gonna just use the agents. That’s just not going to happen,’” Stier said during a separate panel at AI Agenda Live.

But he noted that vibe coding enterprise software will likely prove more useful in replacing legacy enterprise apps for younger startups that “have huge valuations but have 50 times less people.”

Some AI startup founders are similarly skeptical. Kabir Nagrecha, cofounder and CEO of IT software startup Tessera Labs, said he thinks automation will focus more on building agents that can use existing apps like Salesforce or SAP rather than replace them.

“The deal disruption is not at the software level, it's at the user level of that software and rethinking what is the role of that user to interact with an SAP, what is the role of a maintainer of SAP, how do we interact with software, not the software itself,” Nagrecha said during the same panel.

As Anthropic tries to push the boundaries of what AI coding can do, we wouldn’t be surprised to see a similar effort from OpenAI, which has pitched its own AI coding tools as capable of creating anything from websites to internal company apps.

JPMorgan Launches PR Offensive

It’s that time of the cycle again: Big banks are talking publicly about how tech-forward they are, this time with AI.

JPMorgan Chase & Co., for instance, told CNBC it has big plans for using AI to operate more efficiently and get an edge on rivals. Derek Waldron, the bank’s chief data analytics officer, said the bank is using AI to help employees create presentations and complete work projects that involve multiple steps.

Waldron went even further, saying JPMorgan will soon use AI from OpenAI and Anthropic to handle interactions with customers, like helping them find information.

His comments are the latest example of how banks, which used to be reluctant to talk publicly about their adoption of new tech for competitive reasons, are now being more open about using AI in customer-facing situations. Wells Fargo is also using AI from Google in a virtual assistant it provides to customers.

But that doesn’t necessarily mean they’re spending all that much on the technology. When banks like JPMorgan began talking about how much they were embracing cloud-based servers, in reality, they were spending relatively little compared to many other types of cloud customers.

JPMorgan also has a big financial incentive to talk up OpenAI in particular: the bank has committed $10 billion to finance data centers OpenAI will use in Texas.

All of which is to say that Waldron's comment about JPM being “fundamentally rewired” for AI might not mean it will be a heavy spender on the technology in the foreseeable future.

OpenAI on Thursday completed a secondary share sale amounting to $6.6 billion, Bloomberg reported. The sale gives current and prior employees the ability to sell stock at a $500 billion valuation.

The transaction and resulting valuation also grant OpenAI the distinction of now being the world’s most valuable private, venture-backed company, surpassing SpaceX, which was reportedly valued at $400 billion after its own secondary share sale this summer.

According to CNBC, OpenAI had authorized up to $10.3 billion in shares for sale, which was up from its original $6 billion target. But only about two-thirds of what it had authorized ultimately got sold.

Besides OpenAI and SpaceX, several other high-valued private companies have turned to secondary sales, which are often used to reward employees and as a retaining tool for companies not ready to go public.

In February, fintech giant Stripe announced a tender offer in which investors would buy shares from past and present employees at a valuation of $91.5 billion. Last December, Databricks raised $10 billion at a $62 billion valuation in a deal that included a secondary share sale aimed at providing liquidity for current and former employees.

In April, OpenAI also made headlines for an investment of up to $40 billion from Japanese investment conglomerate SoftBank in a deal that marked the largest startup financing ever and valued it at $300 billion.

UK authorities step up security at Jewish sites after what Starmer denounces as a terrorist incident

Two people have been killed in a knife and car attack at a synagogue in Manchester that took place on Yom Kippur, the holiest day in the Jewish calendar, in what Prime Minister Sir Keir Starmer denounced as a terrorist incident.

The assailant in Thursday morning’s attack drove a car into members of the public outside Heaton Park Hebrew Congregation Synagogue and then stabbed people with a knife.

Greater Manchester Police said its officers fatally shot the attacker at 9.38am, seven minutes after being called to the synagogue, as worshippers barricaded themselves inside.

The attacker was wearing a vest that “had the appearance of an explosive device” but was later found to be “not viable”, the police said.

In a televised address to the nation, Starmer described the attacker as a “vile individual who committed a terror attack, who attacked Jews because they are Jews, and attacked Britain because of our values”.

In comments aimed at the Jewish community, he added: “I promise you I will do everything in my power to guarantee you the security you deserve, starting with a more visible police presence protecting your community.”

The prime minister had returned to London early from a summit in Copenhagen to chair the UK government’s Cobra emergency response group.

The Manchester attack is the first fatal incident targeting Jewish people in the UK in at least 30 years, according to data from the Community Security Trust, a UK charity that tracks antisemitism.

However, it comes against the backdrop of a sharp rise in antisemitic incidents in the country in the wake of Hamas’s October 7, 2023 attack on Israel and the subsequent Gaza war. The CST recorded 1,521 incidents in the first half of the year.

In a post on X, Gideon Sa’ar, Israel’s foreign minister, said UK authorities had “failed to take the necessary action to curb this toxic wave of antisemitism and have effectively allowed it to persist”.

Israeli Prime Minister Benjamin Netanyahu later said on X that Israel grieved with the Jewish community in the UK after the attack in Manchester.

“As I warned at the UN: weakness in the face of terrorism only brings more terrorism. Only strength and unity can defeat it,” he added, referring to his speech at the UN General Assembly in New York last week.

UK home secretary Shabana Mahmood said the government had stepped up security at synagogues across the UK.

“We will do whatever is required to keep our Jewish community safe,” she added, speaking at the scene of the Manchester attack on Thursday evening.

Greater Manchester mayor Andy Burnham said the city region “will never let acts that are designed to cause hatred, division in our communities, violence, we will never let them succeed”.

Communities needed to redouble efforts to stand united and work together in the aftermath of the atrocity, he added.

The police believed officers knew the attacker’s identity but did not immediately disclose it. Two arrests have been made.

In addition to the two people killed by the attacker, four remained in hospital after suffering a variety of serious injuries.

Witnesses told police that one of the people attacked was a security guard at the synagogue.

Footage of the scene shared on social media showed the attacker writhing on the ground within the gates of the synagogue when police shot him. An officer shouted to a nearby crowd “he has a bomb” and “move back”.

Shortly after the attack, specialist counter-terrorist police, officers armed with machine guns, and members of the army were seen at the site, as well as a damaged black Kia car near the gates to the synagogue that had been driven off the road.

On Thursday afternoon, a large police cordon was in place in Crumpsall, the Manchester suburb where the attack occurred, which is multicultural with a large Jewish community.

A helicopter flew overhead and members of the public had gathered.

A Jewish woman who lives nearby told the Financial Times that all the nearby synagogues were on “lockdown”, adding the situation was “very worrying”.