AI Coding Blunders and Security Risks: The Case of Netanyahu’s Adviser

Automated Political Monitoring Code Left Public on GitHub

An artificial intelligence program designed to monitor news sources for Prime Minister Benjamin Netanyahu and his political allies was built by longtime adviser Yonatan Urich using the AI tool Claude, with the application’s source code subsequently left publicly accessible on GitHub.

The Tech TL;DR:

• AI-assisted programming tools allowed political aides to rapidly build news-monitoring applications, but unsecured repositories exposed sensitive access tokens and personal identifiers.
• Connecting personal messaging accounts to third-party automation services like Ultramsg bypasses end-to-end transport encryption once messages reach authorized endpoints.

Yonatan Urich Exposed Private Data on GitHub

Every 90 seconds, the AI-assisted program scanned roughly 50 news sources for mentions of Benjamin Netanyahu, Sara Netanyahu, Likud, and political rivals, dispatching summaries to two WhatsApp groups. A September 24 Haaretz investigation revealed that Yonatan Urich left the underlying code publicly available on GitHub. The exposed repository contained private telephone numbers, WhatsApp group identifiers, and access tokens capable of retrieving group data or transmitting messages.

Calcalist subsequently reported that Urich connected his personal WhatsApp account to Ultramsg, a third-party automation service operating from the United Arab Emirates. This connection potentially granted the platform access to unrelated private correspondence. Ultramsg operates via a linked-device protocol. Once authorized, such services receive and process messages after decryption, entirely outside the scope of WhatsApp’s end-to-end transit protection. Urich utilized his personal account rather than a dedicated automation profile, widening the scope of exposed data.

# Conceptual API integration flow for third-party messaging services
import requests

def send_whatsapp_alert(payload, api_token, endpoint_url):
    headers = {
        "Authorization": f"Bearer {api_token}",
        "Content-Type": "application/json"
    }
    response = requests.post(endpoint_url, json=payload, headers=headers)
    return response.status_code

Online Leak of Purported Conversations and Legal Response

On October 11, purported conversations originating from Urich’s WhatsApp account surfaced online. While participants acknowledged select exchanges, the authenticity of the complete dataset remains unverified. Urich’s legal representatives stated that an unauthorized party accessed his communications, asserting that specific messages were altered or supplemented with fabricated material. Lawyers threatened legal action against individuals who obtained or distributed the files.

No public forensic investigation has established a direct causal link between the GitHub code exposure, the Ultramsg integration, and the October 11 leak. Security analysts note that unauthorized access could equally stem from phishing campaigns, device spyware, cloud backups, or alternative endpoints.

Prof. Yossi Oren on the Risks of AI-Assisted Programming

Prof. Yossi Oren, an associate professor at Ben-Gurion University of the Negev’s Institute for Software, Systems and Security, stated that available public information does not indicate an underlying exploit of WhatsApp infrastructure or physical device compromise. “As far as the public knows, there was no hacking of WhatsApp, no hacking of the phone,” Oren told The Media Line. “This is no longer a phone being hacked. The user is hacking himself,” he said regarding the delegation of excessive operational trust to automated systems.

Oren likened the current wave of AI-assisted software generation to early social media adoption, where users published personal media without grasping the scale of distribution. He emphasized that AI coding tools lower technical barriers, allowing non-developers to deploy complex integrations while accidentally leaking credentials and over-privileged API access tokens.

Matias Rott on Political Intelligence and Campaign Vulnerabilities

Cyber intelligence expert Matias Rott noted that political advisers maintain high-value intelligence repositories despite lacking formal cabinet clearances or government positions. “Not being inside the Cabinet meeting does not mean I will not know what happened there,” Rott told The Media Line. Senior figures frequently debrief trusted aides following sensitive policy sessions, sharing strategic perspectives and interpersonal political dynamics.

Rott explained that comprehensive forensic examinations require physical device acquisitions, mirror copies of stored data, and thorough audits of connected third-party services. Israel’s parliamentary election, scheduled for October 27, provides external actors or opportunistic groups a strategic window to weaponize leaked communications. Disrupting a political campaign does not require altering vote tallies; forcing key strategists to spend operational hours verifying documents and managing public relations diverts essential resources from campaigning.

ChatGPT Queries in the Qatargate Investigation and Precedent Incidents

This disclosure follows prior digital breaches targeting Israeli political figures. During the 2019 election campaign, television reports indicated the Shin Bet detected a breach of Benny Gantz’s phone, with suspicions directed toward Iran. In December 2025, the Iran-linked hacking group Handala claimed a successful compromise of former prime minister Naftali Bennett’s phone, publishing personal contacts and files. Bennett’s office acknowledged unauthorized access to his Telegram account, noting the inclusion of fabricated data alongside genuine material.

To mitigate these risks, Rott recommended strict device segregation, advising officials against utilizing personal lines for professional duties. Oren suggested advanced hardening protocols such as Apple’s Lockdown Mode for individuals facing sophisticated digital threats, while stressing the urgent need for comprehensive cybersecurity education across political and journalistic operations.

Disclaimer: The technical analyses and security protocols detailed in this article are for informational purposes only. Always consult with certified IT and cybersecurity professionals before altering enterprise networks or handling sensitive data.