Weekly reading · methodology v0.6
Window ending Oct. 6, 2025
Harms count in full for two weeks after they are reported, then one level less every two weeks. The control floor looks at reports from Sept. 7, 2025 through Oct. 6, 2025. A selected catalogue of reported AI incidents, reviewed through Oct. 7, 2026. This reading describes documented harm in the records or, when none qualifies, the highest control level breached. It is not a forecast or a measure of all AI activity.
Current recalculation
3
Control failures only · No qualifying harm. The reading is the highest control level breached: 3, by 1 record. It is a level, not a count.
Recalculated from the catalogue in this build. Historical backcasts were not published at the time.
Published at the time
No publication snapshot exists for this week.
This is a backcast from the current catalogue. It was not a reading published at the time.
Records behind this reading
4 selected records; 0 documented external harms counting (0 qualifying). Records and ratings below reflect the current catalogue.
No harm counts in this week. The reading is the highest control level breached: 3 (set by 1 record: PI-0029). It is a level, not a count.
- Documented harm
- 0
- No harm found (stated scope)
- 0
- No harm reported
- 3
- Impact unknown
- 0
- Alleged, AI role uncorroborated
- 1
- Counts toward the index
- 0
- Unverified, watching
- 0
- Alleged in court
- 1
- Control failure, tracked
- 3
- Tracked separately
- 0
Documented exclusions: 0 internal; 0 awaiting evidence review. 0 qualifying records with a bounded single-source review.
Inspect the evidence · 4 records
- PI-0028 · Families file three lawsuits alleging Character.AI chatbots harmed teenagers, including a suicideAlleged, AI role uncorroborated · excluded from the harm reading
- PI-0029 · OpenAI and Apollo Research found covert deception, including sandbagging, in o3 and o4-mini evaluationsNo harm reported · sets the reading: highest control level breached
- PI-0027 · ShadowLeak: hidden email text made ChatGPT's Deep Research agent send inbox data from OpenAI's serversNo harm reported · excluded from the harm reading
- PI-0030 · ForcedLeak: web form submissions could make Salesforce Agentforce leak CRM lead dataNo harm reported · excluded from the harm reading
Status totals cover counted records; aliases, superseded aggregates and records tracked separately are excluded. Eligibility and extent notes overlap those totals. These observations are not statistical uncertainty bounds.
- PI-0028Families file three lawsuits alleging Character.AI chatbots harmed teenagers, including a suicideSept. 16, 2025 · Severe harm, alleged · Deployment
- PI-0029OpenAI and Apollo Research found covert deception, including sandbagging, in o3 and o4-mini evaluationsSept. 17, 2025 · No harm reported · Controlled test
- PI-0027ShadowLeak: hidden email text made ChatGPT's Deep Research agent send inbox data from OpenAI's serversSept. 18, 2025 · No harm reported · Controlled test
- PI-0030ForcedLeak: web form submissions could make Salesforce Agentforce leak CRM lead dataSept. 25, 2025 · No harm reported · Controlled test
Sources
These links support the records above. Source availability and conclusions may change.
- Social Media Victims Law Center: Social Media Victims Law Center Files Three New Lawsuits on Behalf of Children Who Died of Suicide or Suffered Sex Abuse by Character.AI (archived copy)Cited in PI-0028
- U.S. District Court, D. Colorado (via CourtListener RECAP): Montoya and Peralta v. Character Technologies et al.: complaintCited in PI-0028
- U.S. District Court, N.D. New York (via CourtListener RECAP): P.J. v. Character Technologies et al.: complaintCited in PI-0028
- CourtListener (RECAP): E.S. v. Character Technologies, Inc. (D. Colo. 1:25-cv-02906): docketCited in PI-0028
- CourtListener (RECAP): Montoya v. Character Technologies, Inc. (D. Colo. 1:25-cv-02907): docketCited in PI-0028
- CourtListener (RECAP): P.J. v. Character Technologies, Inc. (N.D.N.Y. 1:25-cv-01295): docketCited in PI-0028
- U.S. District Court, D. Colorado (via CourtListener RECAP): E.S. v. Character Technologies et al.: complaintCited in PI-0028
- Bloomberg Law: Character.AI, Google agree to settle teen chatbot harm lawsuitsCited in PI-0028
- CNN: More families sue Character.AI developer, alleging app played a role in teens' suicide and suicide attemptCited in PI-0028
- CNN: Character.AI and Google agree to settle lawsuits over teen mental health harms and suicidesCited in PI-0028
- CNN (via KVIA): More families sue Character.AI developer, alleging app played a role in teens' suicide and suicide attemptCited in PI-0028
- CNN (via KTVZ): Character.AI and Google agree to settle lawsuits over teen mental health harms and suicidesCited in PI-0028
- OpenAI: Detecting and reducing scheming in AI modelsCited in PI-0029
- Apollo Research: Stress Testing Deliberative Alignment for Anti-Scheming TrainingCited in PI-0029
- arXiv (Schoen et al., Apollo Research and OpenAI): Stress Testing Deliberative Alignment for Anti-Scheming TrainingCited in PI-0029
- eWeek: AI Might 'Scheme' Less Thanks to This OpenAI ResearchCited in PI-0029
- Open Data Science: New research highlights scheming risks in AI modelsCited in PI-0029
- Radware: ShadowLeak: A Zero-Click, Service-Side Attack Exfiltrating Sensitive Data Using ChatGPT's Deep Research AgentCited in PI-0027
- The Hacker News: ShadowLeak zero-click flaw leaks GmailCited in PI-0027
- The Record: OpenAI fixes zero-click ShadowLeak vulnerabilityCited in PI-0027
- Noma Security: ForcedLeak: AI agent risks exposed in Salesforce AgentforceCited in PI-0030
- The Register: Prompt injection – and a $5 domain – trick Salesforce Agentforce into leaking salesCited in PI-0030
- The Hacker News: Salesforce patches critical ForcedLeakCited in PI-0030
- CSO Online: Vulnerability in Salesforce AI could be tricked into leaking CRM dataCited in PI-0030
All weekly readings · How this reading is calculated · Download the weekly card
