Ongeautoriseerde toegang tot Anthropic's Claude Mythos AI via vendorlek
Anthropic, het bedrijf achter Claude Mythos AI, onderzoekt ongeautoriseerde toegang tot zijn AI-model via de omgeving van een externe leverancier. Dit incident vond plaats slechts twee weken na de introductie van Mythos. Een kleine groep gebruikers op een Discord-kanaal heeft naar verluidt toegang verkregen tot Mythos, met als primair doel het verzamelen van gegevens.
Signal context
First seen: Apr 22, 2026
Last updated: Jul 3, 2026
Status: Public signal
Key points
- Ongeautoriseerde toegang tot Anthropic's Claude Mythos AI-model.
- Toegang verkregen via een externe leveranciersomgeving.
- Incident vond plaats twee weken na de introductie van Mythos.
Signal analysis
BetaIt helps compare this signal with other published signals without treating the labels as final determinations.
Likely country: 🇳🇱 Netherlands
inferred from source domains
Watch internet-facing systems, credential abuse and exploit activity.
- Source type: outside the affected organization
Impact area: Confidentiality
- 1 signal in the same sector
- 89 signals with the same likely impact area
- 1 signal linked to this organization/domain
External sources
Cyberaanvallen april 2026 in Nederland en België - Cybercrimeinfohttps://www.ccinfo.nl/menu-nieuws-trends/actuele-cyberaanvallen/cyberaanvallen-2026-aprilPublic source from ccinfo.nl.
Beursnieuwsgeschiedenis per 22-04-2026 | MarketScreener ...https://nl.marketscreener.com/beursnieuws/historie/2026/22-april/?p=9Public source from nl.marketscreener.com.
Related signals
Grouped by why the signal is relevant.
Singapore Land Authority Data Breach Exposes 70,000 Records via IBM Testing Environment
Personal data belonging to about 70,000 individuals has been compromised in a cybersecurity incident involving the Singapore Land Authority (SLA) and a cloud environment managed by IBM. The breach stemmed from unauthorized access to a dataset created for vendor development and systems integration testing, which was intended to contain only mock and anonymized data. However, it was found to contain names, NRIC (National Registration Identity Card) numbers, and property addresses. IBM has revoked access, and SLA is notifying affected individuals and investigating with government agencies.
CTIVD: AIVD en MIVD verwerken persoonsgegevens in bulkdata onrechtmatig
The Dutch intelligence services, AIVD and MIVD, have unlawfully processed personal data in bulk datasets, according to a ruling by the Committee for the Supervision of the Intelligence and Security Services (CTIVD). The report, published on July 1, 2026, states that groups of employees had unauthorized access to personal data, and large quantities of data were stored for too long. The bulk datasets, sometimes containing millions of records, include names, phone numbers, location data, social media data, and communication content, sourced from government agencies, commercially available datasets, or stolen datasets offered by criminals. The CTIVD has issued thirteen recommendations to improve the situation.
Aflac Life Insurance Japan Suffers Cybersecurity Breach Exposing Policyholder Data
Aflac Life Insurance Japan disclosed unauthorized access to its systems between June 15 and June 25, 2026. The breach affected files containing policy details, personal information, and bank account information of approximately 4.38 million customers. The company has suspended affected systems and is investigating the incident with third-party cybersecurity experts.
Polymarket Confirms Supply Chain Attack, $3 Million Stolen
Polymarket, a cryptocurrency-based prediction market, confirmed a supply chain attack. A breach of a third-party frontend vendor led to malicious JavaScript being injected into its website. Attackers tricked users into approving fraudulent transactions, resulting in approximately $3 million being stolen from fewer than 15 accounts. The backend systems remained unaffected.
National Association of Insurance Commissioners (NAIC) Confirms Data Breach via Oracle PeopleSoft Zero-Day
The National Association of Insurance Commissioners (NAIC), a US insurance regulatory standards body, confirmed a cyberattack after the ShinyHunters group claimed theft of 3.1TB of data. The breach was reportedly achieved through an Oracle PeopleSoft zero-day vulnerability. ShinyHunters claimed access to regulatory filings, production logs, cloud configuration files, and other internal records.
JCOM Co., Ltd. Affected by KDDI Corporation Data Breach
JCOM Co., Ltd., a Japanese internet service provider, was impacted by a data breach originating from an email system provided by KDDI Corporation. Threat actors gained unauthorized access to this shared system by exploiting a vulnerability in third-party software. This led to the potential exposure of up to 14.2 million email addresses and passwords across all affected ISPs. JCOM customers' email addresses and passwords may have been compromised.