Anthropic's Deputy Chief Information Security Officer Jason D. Clinton has warned that digital threats could worsen over the next 18 months.
In a post on X, he said attackers are using existing artificial intelligence capabilities to identify and exploit vulnerabilities before defenders can respond.
"As bad as things have gotten in the past 4 days, they are going to get a lot worse before they get better," Clinton said.
He urged security professionals to seek greater support from corporate leadership and to prioritize fundamental protections rather than focusing exclusively on artificial intelligence.
Clinton also referenced an earlier warning from August about a coming surge in software vulnerabilities and the difficulty of patching critical infrastructure quickly enough.
AI Tools Could Give Attackers an Edge
Clinton has highlighted the concept of "vulnzero," referring to zero vulnerabilities.
His outlook centers on the prospect that increasingly capable AI tools could eventually help developers identify and eliminate software weaknesses before applications enter production.
Until those defensive tools become widely available, however, existing AI capabilities could give attackers an advantage.
Automated vulnerability discovery could help them identify security flaws in older systems, exploit known weaknesses more rapidly, and act before organizations deploy corrective patches.
Clinton's warning has drawn criticism from people who argue that AI companies are contributing to escalating cyber risks before adequate defenses are in place.
His proposed timeline calls for the industry to withstand an estimated 18-month period before more accessible defensive technologies become widely adopted.
The warning follows a series of incidents affecting organizations across Europe and major online services.
According to a report published by Threema on October 19, attacks compromised millions of personal records, exposed customer information, and disrupted supermarket refrigeration systems.
Check Point recorded 48% more attack attempts worldwide in September 2026 than in September 2025, according to Threema.
Expanded reporting requirements under European Union frameworks, including NIS2 and DORA, may also contribute to increased public awareness of cyber incidents.
Breaches Expose Data Across Europe
Threema's review of incidents from late September through mid-October documented several significant breaches in Switzerland and elsewhere.
Publica, a federal pension fund, and the City of Zurich's pension fund were affected by an attack targeting software vendor PK Softech.
The breach potentially exposed salaries, retirement savings, account numbers, addresses, social security numbers, and spousal information.
Swiss space company Beyond Gravity also suffered a targeted attack that appeared to involve data theft.
Media reports cited suspicions of Chinese involvement, but no official confirmation established who was responsible.
At online marketplace Ricardo, attackers exploited a security vulnerability and accessed names, addresses, and telephone numbers associated with approximately 890,000 accounts.
Email addresses and passwords were not affected, according to Threema.
Some customers subsequently received personalized phishing messages through WhatsApp, although a connection between those messages and the breach remained unconfirmed.
Swiss online magazine Republik disclosed that attackers exploited an unpatched vulnerability in the Metabase analytics tool.
A fix had been available since August 6, 2026.
The compromised information included names and email addresses of current and former readers, with some records also containing telephone numbers, birth years, and genders.
Credit card information was not compromised.
Swiss public broadcaster SRG reported that data belonging to approximately 340 current and former employees had been accessed.
The information included names, job titles, and contact details from 2020, some of which were private.
SRG said there was no evidence that the information had been published or that passwords, financial data, or journalistic sources had been stolen.
In Denmark, attackers stole approximately 8.8 million records from the central population register through a legitimate interface operated by a private IT company.
At least three company accounts reportedly used the password "123456."
The compromised records contained information associated with the CPR number, a widely used identifier for government agencies, banks, and hospitals.
Other incidents involved British online retailer ASOS and communications platform Discord.
According to Threema, attackers used social engineering to obtain an ASOS employee's credentials and accessed third-party platforms, compromising customer information and search histories.
In the Discord-related incident, attackers targeted Double Counter, an external security service, rather than Discord itself, potentially exposing identifiers, usernames, IP addresses, and email addresses.
Several French supermarket chains also experienced disruptions after attackers manipulated refrigeration and freezer systems into defrost mode, forcing stores to discard large quantities of food.
The attack targeted a Danish software provider serving the affected chains.
Separately, attackers impersonating law enforcement through an Italian government agency's official communications channel persuaded British online bank Revolut to disclose information about nearly 700 customers.
The data included passport details, account numbers, and addresses.
No money was stolen, and media reports indicated that most affected customers were in Switzerland.
Threema identified artificial intelligence, expanding digital services, and increasingly complex technology supply chains as factors that may contribute to the threat landscape.
Supply chain attacks exploit weaknesses in third-party providers to reach organizations that depend on their software or services, as illustrated by the incidents involving Publica and Republik.
