Meta, OpenAI, Anthropic ‘rogue’ AI hacks all sourced to Israeli intelligence operatives

One Tel Aviv-based tech ‘security’ company, Irregular, was at the centre of “rogue” hack attacks launched during AI testing procedures. Meta, OpenAI (of ChatGPT), and Anthropic’s AI models were all involved.

Reporting at the time, the Wall Street Journal ran with this alarming headline:

Over the course of just two weeks, OpenAI, Anthropic and Meta all revealed that their AI models went “rogue” during what they called “routine security testing“. And Irregular, a “small Israeli startup,” was behind it.

CNBC underscored the risks posed by unguarded AI models and incidents like this:

With the leading models becoming ever more powerful, their ability to act in malicious ways is turning into a major threat for corporations and governments, especially as the risk involves hacking into critical computer systems and infrastructure. The recent exploits at OpenAI, Anthropic and Meta all involved their AI models accessing websites that should have been off-limits as part of the cybersecurity testing.

One Israeli tech company, Irregular, was repeatedly identified as hosting the so-called “evaluation testbed.” It’s that technology platform which effectively enabled consumer-facing AI models to turn into hack-bots.

This began by allowing OpenAI models and Anthropic’s Claude to access the internet. Meta stated that they do not yet have “all the facts.” Meanwhile the firm, Irregular, stated to CNBC that the hacking incident:

did not involve a sandbox escape or a sophisticated cyber action … [adding] there are no current open issues.

A ‘cyber offensive’ Zionist project

The small tech startup is valued at $450m, backed by $80m from Sequoia Capital and Redpoint Ventures. Sequoia’s partner, Shaun Maguire, is a known Zionist and financial backer of Israel’s genocidal wars and illegal occupations. He actually relaunched Sequoia into Israel during its murderous Gaza onslaught.

Sequoia is a key target of pro-Palestinian boycotts (BDS). This is because recent headlines read like:

Sequoia bets on Israel because of the war

Maguire previously called Israel’s booming war economy “next-level” and has said that:

Maintaining the world’s most advanced cyber capabilities is existential for Israel.

Maguire and his partner at Sequoia, Dean Meyer, wrote in a blog post that the team at Irregular is:

able to see around corners others can’t, running cyber offensive evaluations on advanced models and developing defenses before those models are released.

Erica Brescia, a managing director at Redpoint Ventures, said publicly as the Gaza genocide unfolded in October 2023 that the VC firm

would give money to [Israeli] first responders in the region.

Redpoint has a track record of investing in tech companies funded by Israel’s Occupation Forces, or IOF. These include Dux Security, which secured a $9 million seed round in late 2025 led by Redpoint and TLV Partners. Its founders were graduates of the IOF’s elite Talpiot program, otherwise labelled:

the IDF’s Super-Secret Elite Brain Trust

Sequoia and Redpoint are co-lead investors in another Israeli “autonomous AI systems” tech company, Glow. Glow was initially valued at $1.2bn, making it a tech “unicorn” and highly valuable. Not masking its corporate imperialism, Redpoint also invested $28m in an IOF intelligence veteran-led AI company “imper.ai”.

Most striking about Irregular, however, is that it’s led by Dan Lahav and Omer Nevo. Both served in different notorious intelligence units of the IOF; Nevo for 11 years. Israel news outlet, Ynet, says of Lahav and Nevo:

both enlisted in the IDF intelligence units, Lahav to Unit 81 in Military Intelligence, Nevo to Unit 8200. Despite the well-known rivalry between the two units, it seems the two live in wonderful harmony.

Zionist AI spy corps

Microsoft reportedly cut Unit 8200 off from using its cloud servers following a major exposé of the spy corps in 2025. Investigative magazine +972 revealed that it has been illegally intercepting and storing the private telephone communications of Palestinian civilians en-masse.

But even as far back as 2014, the Guardian reported that officials in the Zionist intelligence corps have “no boundaries.” That article cites one Intelligence Corps employee as saying:

I did and encountered things that seemed irrelevant from a security standpoint, and I did not have a clear conscience participating in such activities. Contrary to my expectations, our database included not only security-related intelligence but also personal and political information. That is to say, on a personal level, there is no respect for Palestinian privacy.

From a political standpoint, information is collected that can serve to manipulate Israeli, Palestinian and international politics. … it has serious impact on the lives of many people …

One ex-operative was quoted, understandably anonymous but verbatim, as admitting:

I realised that the job I had done during my military service was that of the oppressor.

Another intelligence hack said, albeit more glibly:

We take advantage of the impact that we have on their lives. Sometimes it involves truly harming a person’s life, or their soul. … It can really screw up their lives. It made me feel omnipotent.

Yet another interviewee admitted to targeting entirely “innocent” Palestinian civilians:

[I] collected information on people who were completely innocent, and whose only crime was that they interested the Israeli security system for various reasons. For reasons they had absolutely no way of knowing. All Palestinians are exposed to non-stop monitoring without any legal protection.

This palpably illegal, draconian behaviour is reprehensible to many — even Israelis. Back in 2014, dozens of veterans from the occupation spy corps condemned its behaviour and refused to participate anon.

Featured image via the Canary 

By Cameron Baillie