Ousted Savannah police officer gave ICE agent husband access to Flock system
TL;DR Highlight
A police officer shared her Flock login with her ICE-agent husband; an audit found 983 searches and she was fired
Who Should Read
Backend and security developers responsible for account sharing and audit-log design in internal SaaS tools
Core Mechanics
- One shared login led to 983 searches
- An internal audit flagged unusually high search volume on one account between January and July 2026.
- The audit found 983 search entries on 97 license plates from Georgia, New Jersey, Maryland, Florida, South Carolina and North Carolina.
- The husband had no account of his own
- He is an ICE agent but had no Flock access through his federal job.
- He said he used his wife's username and password for about four months.
- The audit log named only the wife
- Searches were recorded under her account, so the system identified her as the user.
- The department concluded this undermined its ability to oversee use of sensitive information.
- Discipline rested on internal policy, not law
- The investigation did not find a violation of Georgia's automated license plate reader law.
- The case centered on a department information-management policy that bars unauthorized people from accessing software.
Community Reactions
- Opinion: shared logins are riskier on SaaS
- Commenters questioned why the system is standard SaaS rather than locked to department devices.
- Another comment guessed the system lacked 2FA.
- Suggestion to restrict access physically
- One commenter said such systems should only be accessible inside secured facilities with badge entry.
- The argument is that this would reduce avenues for abuse.
- Opinion: authorized users still abuse access
- Commenters argued that people with approved access abuse surveillance tools again and again.
- One cited Snowden saying NSA workers used surveillance systems to peep on partners and exes.
- Opinion: blame the vendor too
- One commenter found it odd that officers get blamed but Flock, the supplier, does not.
- They argued Flock is built to do more than read plates.
How to Apply
- Find shared logins in your internal tools
- Today, run a log query for accounts seen from many IPs, devices or regions in a short window.
- Confirm with each owner and issue individual accounts.
- Add 2FA and device limits to sensitive systems
- Require SSO and 2FA on internal tools that can look up or track people.
- Where possible, restrict logins to managed devices or allowed IPs.
- Create a separate account process for outside agencies
- Make other teams or agencies request an account in their own name.
- When someone searches on another's behalf, log the requester and purpose.
- Add alerts for abnormal search volume
- Set per-user daily and weekly search baselines and notify an auditor when exceeded.
- Include the lookup target and a purpose field in the alert.
Try It
# Paste into Claude or ChatGPT (account-sharing audit prompt based on the article's case)
I'm a backend developer reviewing access logs of an internal lookup system.
Find patterns in the logs below that suggest one account is shared by multiple people.
Logs:
{{access log sample (user_id, timestamp, IP, User-Agent, search target, count)}}
Please do the following.
1. Tabulate accounts with varied IPs, devices or regions in a short period.
2. Flag accounts whose search count spiked versus their usual volume.
3. Write 3 questions to verify for each suspicious account.
Next, try asking: what checks become easier if I add a 'purpose of search' field to the logs?Terminology
Related Papers
Same Simulation, Different Evidence: Where Retrieved Tool Output Costs LLM Agents Accuracy
Saving long tool output to files cut GPT-5.1 accuracy by 11.3 points; naming the authoritative file restored it.
Fake News, Real Crashes: Information Injection Attacks on LLM Financial Trading Agents
Fake headlines hijacked LLM trading agents 62% of the time versus 14% for prompt injection.
Harness Engineering 101: The LLM is the brain, the harness is the body
An agent is a loop managing a JSON array; 16 chapters build one from scratch
Harness2 by Google Research – Recursive Agent Harnessing
Recursively refining the harness with fixed model weights raised benchmark scores by up to 11.8 points.
Tmux and Claude Code: Sessions, Send-Keys, and the Ceiling
Running Claude Code agents in tmux works well, but limits appear around four agents.
Hybrid Reasoning Agents: LLM agents that think in latent space, act through text
Mixing latent reasoning with text tool use cut tokens by up to 54.7% and lifted performance 10.4%.