THE RUMP REPORT

ANTHROPIC RESEARCHER QUITS OVER AI EXTINCTION FEARS... DEMANDS PACING PACT BETWEEN LABS

techcrunch-ai · SEP 9

REPORT: EX-ANTHROPIC RESEARCHER REVEALS FIRM'S 'MINI MANHATTAN PROJECT' ON SELF-IMPROVING AI

techmeme · SEP 10

ANTHROPIC RESEARCHER QUITS... WARNS IT'S 'CRUNCH TIME FOR HUMANITY'

wired-ai · SEP 9

META'S DEFENSE: CHILD ABUSE ADS AREN'T SAN FRANCISCO'S PROBLEM...

wired-ai · SEP 9

HAWLEY'S SENATE DISASTER PANEL PROBES OPENAI OVER HUGGING FACE BREACH... CALLS IT 'RECKLESS'

techmeme · SEP 10

OPINION: TECHCRUNCH COLUMNIST SAYS OPENAI'S HUGGING FACE BREACH SHOWS DANGERS OF SUPERHUMAN AI

techcrunch-ai · SEP 9

OPENAI CLAIMS IT CRACKED $1M MATH PROBLEM... THEN ACCUSED OF STEALING CREDIT

mit-tech-review · SEP 9

HE STRIPPED AN AI'S GUARDRAILS... IT HACKED HIS OWN PC

wired-ai · SEP 9

SUNO SAYS NEW AI MODEL DROPS OLD TRAINING DATA AS COPYRIGHT LAWSUITS PILE UP...

techcrunch-ai · SEP 9

NEW APPLE WATCH AI CAN TRANSCRIBE YOUR SPEECH, SUMMARIZE CONVERSATIONS... APPLE SAYS NO RAW AUDIO SAVED

techcrunch-ai · SEP 9

APPLE WATCH'S NEW AI FEATURE LISTENS ALL DAY... SUMMARIZES YOUR CONVERSATIONS

hn-frontpage · SEP 10

REPORT: JULY TRANSMISSION FAULT KNOCKED 3+ GIGAWATTS OFF VIRGINIA'S AI DATA CENTER GRID IN SECONDS

mit-tech-review · SEP 10

MICROSOFT SIGNS ENFORCEABLE AI PACT WITH TEACHERS UNION: NO TRAINING ON STUDENT DATA, NO AI 'COMPANIONS'

verge-ai · SEP 9

OPENAI CALLS ASTRA 'MOST ALIGNED AVAILABLE MODEL'... STAFFER: METRIC 'WILL BREAK DISCONTINUOUSLY'

zvi · SEP 9

SOURCES: CHINA'S AI CHIP GIANTS JACK PRICES UP TO 50%... HBM COSTS SURGE

techmeme · SEP 10

SECOND MATHEMATICIAN ACCUSES OPENAI OF 'DISHONESTY' OVER TRAINING DATA TRANSPARENCY...

verge-ai · SEP 10

[PAPER] STUDY: NEITHER SIDE OF THE AI WATERMARK FIGHT CAN PROVE ITS CASE... CALLS UNVERIFIABILITY THE REAL FAILURE

arxiv · SEP 9

[PAPER] STUDY: ONE HACKED AI VENDOR COULD CASCADE INTO FULL-BLOWN BANKING CRISIS

arxiv · SEP 9

[PAPER] REPORT: FINE-TUNING GUTS AI SAFETY ENFORCEMENT... OUTSIDE CONTROLS CAN SURVIVE

arxiv · SEP 9

[PAPER] STUDY: MIXED-ATTACK TRAINING DELIVERS MOST BALANCED DEFENSE FOR AI LOAN SCORING

arxiv · SEP 9

[PAPER] STUDY: 'PRIVATE' AI SYNTHETIC DATA STILL LEAKS WHO WAS IN IT... TOP 10% OF RECORDS CARRY 40% OF THE RISK

arxiv · SEP 9

[PAPER] STUDY: FAKE 'FICTIONAL SCENARIO' TRICKS AI INTO WRITING MALWARE... GUARDRAILS FAIL NEARLY 100% OF TIME

arxiv · SEP 9

[PAPER] REPORT: VOICE-CLONE SCAMS NOW SPLICE FAKE SENTENCES INTO REAL CALLS...

arxiv · SEP 9

[PAPER] STUDY: NEW BENCHMARK ASKS THE QUESTION — CAN AI BUILD THE SYSTEMS THAT RUN IT?

arxiv · SEP 9

[PAPER] STUDY: RESEARCHERS UNVEIL 'A-JIT' — SOFTWARE THAT NEVER STOPS REWRITING ITSELF...

arxiv · SEP 9

[PAPER] STUDY: AI MODELS CAN'T PREDICT THEIR OWN BEHAVIOR... SELF-REPORTS BARELY BEAT RANDOM CHANCE

arxiv · SEP 9