← Voltar ao feed

Sponsored piece: carriers retain call logs for years, citing 2024 AT&T breach of 109M customers' records

1 fonte(s) · 1 atualização(ões) · publicado 02/11/2026 12:01 · coletado há 2 horas

This is a Cape-sponsored article, not a new incident report. It argues that long retention of call metadata by mobile carriers is a risk in itself, and it cites the April 2024 AT&T breach, which exposed communication logs of about 109 million current and former customers. The exposed data covered May to October 2022 and January 2023: phone numbers, call and text counts, call durations and location data. AT&T said it did not include call content, customer names or Social Security numbers. The article also recalls that AT&T was one of several Snowflake customers hit by credential-based thefts, and that Salt Typhoon reportedly compromised every major US carrier. It then promotes Cape's short-retention model.

Detalhes técnicos

  • Carriers generate SIP messages (session setup and teardown, used for troubleshooting) and Call Detail Records (CDRs) (who, when, how long, often which cell site, used for billing and roaming settlement).
  • Metadata can reveal close relationships, home and work locations and daily routines, and can be correlated with public data to derive identities.
  • In 2022 carriers told the FCC chair that they kept geolocation data for up to five years.
  • Cape says it keeps SIP messages for 14 days, deletes internal CDRs after 1 day, keeps daily totals for up to 60 days and holds partner-supplied CDRs for 30 days. It also rotates network identifiers.
  • Cape says Trail of Bits audited the feature in May and July 2026 and found no evidence that the two retention claims were violated. This is a vendor claim.

Remediação

  • Reduce exposure by minimizing retained metadata. The article recommends carriers with short retention windows.
  • High-risk users (journalists, executives, security staff) should treat carrier retention policy as part of their threat model.

Linha do tempo

  1. bleepingcomputer.com ↗ · pub. 02/11/2026 12:01 · coletado 02/11 13:00 primeiro registro

    Sponsored piece: carriers retain call logs for years, citing 2024 AT&T breach of 109M customers' records

    ler a análise desta fonte

    This is a Cape-sponsored article, not a new incident report. It argues that long retention of call metadata by mobile carriers is a risk in itself, and it cites the April 2024 AT&T breach, which exposed communication logs of about 109 million current and former customers. The exposed data covered May to October 2022 and January 2023: phone numbers, call and text counts, call durations and location data. AT&T said it did not include call content, customer names or Social Security numbers. The article also recalls that AT&T was one of several Snowflake customers hit by credential-based thefts, and that Salt Typhoon reportedly compromised every major US carrier. It then promotes Cape's short-retention model.

    Technical Details

    • Carriers generate SIP messages (session setup and teardown, used for troubleshooting) and Call Detail Records (CDRs) (who, when, how long, often which cell site, used for billing and roaming settlement).
    • Metadata can reveal close relationships, home and work locations and daily routines, and can be correlated with public data to derive identities.
    • In 2022 carriers told the FCC chair that they kept geolocation data for up to five years.
    • Cape says it keeps SIP messages for 14 days, deletes internal CDRs after 1 day, keeps daily totals for up to 60 days and holds partner-supplied CDRs for 30 days. It also rotates network identifiers.
    • Cape says Trail of Bits audited the feature in May and July 2026 and found no evidence that the two retention claims were violated. This is a vendor claim.

    Remediation

    • Reduce exposure by minimizing retained metadata. The article recommends carriers with short retention windows.
    • High-risk users (journalists, executives, security staff) should treat carrier retention policy as part of their threat model.